117LDAPMAP Lightweight Directory Access Protocol support. You will
118 have to install the UMich or OpenLDAP
119 (http://www.openldap.org/) ldap and lber libraries to use
120 this flag.
121MAP_REGEX Regular Expression support. You will need to use an
122 operating system which comes with the POSIX regex()
123 routines or install a regexp library such as libregex from
124 the Free Software Foundation.
125PH_MAP PH map support. You will need the qi PH package.
126MAP_NSD nsd map support (IRIX 6.5 and later).
127
128>>> NOTE WELL for NEWDB support: If you want to get ndbm support, for
129>>> Berkeley DB versions under 2.0, it is CRITICAL that you remove
130>>> ndbm.o from libdb.a before you install it and DO NOT install ndbm.h;
131>>> for Berkeley DB versions 2.0 through 2.3.14, remove dbm.o from libdb.a
132>>> before you install it. If you don't delete these, there is absolutely
133>>> no point to including -DNDBM, since it will just get you another
134>>> (inferior) API to the same format database. These files OVERRIDE
135>>> calls to ndbm routines -- in particular, if you leave ndbm.h in,
136>>> you can find yourself using the new db package even if you don't
137>>> define NEWDB. Berkeley DB versions later than 2.3.14 do not need
138>>> to be modified. Please also consult the README in the top level
139>>> directory of the sendmail distribution for other important information.
140>>>
141>>> Further note: DO NOT remove your existing /usr/include/ndbm.h --
142>>> you need that one. But do not install an updated ndbm.h in
143>>> /usr/include, /usr/local/include, or anywhere else.
144
145If NEWDB and NDBM are defined (but not NIS), then sendmail will read
146NDBM format alias files, but the next time a newaliases is run the
147format will be converted to NEWDB; that format will be used forever
148more. This is intended as a transition feature.
149
150If NEWDB, NDBM, and NIS are all defined and the name of the file includes
151the string "/yp/", sendmail will rebuild BOTH the NEWDB and NDBM format
152alias files. However, it will only read the NEWDB file; the NDBM format
153file is used only by the NIS subsystem. This is needed because the NIS
154maps on an NIS server are built directly from the NDBM files.
155
156If NDBM and NIS are defined (regardless of the definition of NEWDB),
157and the filename includes the string "/yp/", sendmail adds the special
158tokens "YP_LAST_MODIFIED" and "YP_MASTER_NAME", both of which are
159required if the NDBM file is to be used as an NIS map.
160
161All of these flags are normally defined in the DBMDEF line in the
162Makefile.
163
164If you define NEWDB or HESIOD you get the User Database (USERDB)
165automatically. Generally you do want to have NEWDB for it to do
166anything interesting. See above for getting the Berkeley DB
167package (i.e., NEWDB). There is no separate "user database"
168package -- don't bother searching for it on the net.
169
170Hesiod and LDAP require libraries that may not be installed with your
171system. These are outside of my ability to provide support. See the
172"Quirks" section for more information.
173
174The regex map can be used to see if an address matches a certain regular
175expression. For example, all-numerics local parts are common spam
176addresses, so "^[0-9]+$" would match this. By using such a map in a
177check_* rule-set, you can block a certain range of addresses that would
178otherwise be considered valid.
179
180
181+---------------+
182| COMPILE FLAGS |
183+---------------+
184
185Wherever possible, I try to make sendmail pull in the correct
186compilation options needed to compile on various environments based on
187automatically defined symbols. Some machines don't seem to have useful
188symbols available, requiring that a compilation flag be defined in
189the Makefile; see the devtools/OS subdirectory for the supported
190architectures.
191
192If you are a system to which sendmail has already been ported you
193should not have to touch the following symbols. But if you are porting,
194you may have to tweak the following compilation flags in conf.h in order
195to get it to compile and link properly:
196
197SYSTEM5 Adjust for System V (not necessarily Release 4).
198SYS5SIGNALS Use System V signal semantics -- the signal handler
199 is automatically dropped when the signal is caught.
200 If this is not set, use POSIX/BSD semantics, where the
201 signal handler stays in force until an exec or an
202 explicit delete. Implied by SYSTEM5.
203SYS5SETPGRP Use System V setpgrp() semantics. Implied by SYSTEM5.
204HASFCHMOD Define this to one if you have the fchmod(2) system call.
205 This improves security.
206HASFCHOWN Define this to one if you have the fchown(2) system call.
207 This is required for the TrustedUser option.
208HASFLOCK Set this if you prefer to use the flock(2) system call
209 rather than using fcntl-based locking. Fcntl locking
210 has some semantic gotchas, but many vendor systems
211 also interface it to lockd(8) to do NFS-style locking.
212 Unfortunately, may vendors implementations of fcntl locking
213 is just plain broken (e.g., locks are never released,
214 causing your sendmail to deadlock; when the kernel runs
215 out of locks your system crashes). For this reason, I
216 recommend always defining this unless you are absolutely
217 certain that your fcntl locking implementation really works.
218HASUNAME Set if you have the "uname" system call. Implied by
219 SYSTEM5.
220HASUNSETENV Define this if your system library has the "unsetenv"
221 subroutine.
222HASSETSID Define this if you have the setsid(2) system call. This
223 is implied if your system appears to be POSIX compliant.
224HASINITGROUPS Define this if you have the initgroups(3) routine.
225HASSETVBUF Define this if you have the setvbuf(3) library call.
226 If you don't, setlinebuf will be used instead. This
227 defaults on if your compiler defines __STDC__.
228HASSETREUID Define this if you have setreuid(2) ***AND*** root can
229 use setreuid to change to an arbitrary user. This second
230 condition is not satisfied on AIX 3.x. You may find that
231 your system has setresuid(2), (for example, on HP-UX) in
232 which case you will also have to #define setreuid(r, e)
233 to be the appropriate call. Some systems (such as Solaris)
234 have a compatibility routine that doesn't work properly,
235 but may have "saved user ids" properly implemented so you
236 can ``#define setreuid(r, e) seteuid(e)'' and have it work.
237 The important thing is that you have a call that will set
238 the effective uid independently of the real or saved uid
239 and be able to set the effective uid back again when done.
240 There's a test program in ../test/t_setreuid.c that will
241 try things on your system. Setting this improves the
242 security, since sendmail doesn't have to read .forward
243 and :include: files as root. There are certain attacks
244 that may be unpreventable without this call.
245USESETEUID Define this to 1 if you have a seteuid(2) system call that
246 will allow root to set only the effective user id to an
247 arbitrary value ***AND*** you have saved user ids. This is
248 preferable to HASSETREUID if these conditions are fulfilled.
249 These are the semantics of the to-be-released revision of
250 Posix.1. The test program ../test/t_seteuid.c will try
251 this out on your system. If you define both HASSETREUID
252 and USESETEUID, the former is ignored.
253HASLSTAT Define this if you have symbolic links (and thus the
254 lstat(2) system call). This improves security. Unlike
255 most other options, this one is on by default, so you
256 need to #undef it in conf.h if you don't have symbolic
257 links (these days everyone does).
258HASSETRLIMIT Define this to 1 if you have the setrlimit(2) syscall.
259 You can define it to 0 to force it off. It is assumed
260 if you are running a BSD-like system.
261HASULIMIT Define this if you have the ulimit(2) syscall (System V
262 style systems). HASSETRLIMIT overrides, as it is more
263 general.
264HASWAITPID Define this if you have the waitpid(2) syscall.
265HASGETDTABLESIZE
266 Define this if you have the getdtablesize(2) syscall.
267HAS_ST_GEN Define this to 1 if your system has the st_gen field in
268 the stat structure (see stat(2)).
269HASSRANDOMDEV Define this if your system has the srandomdev(3) function
270 call.
271HASURANDOMDEV Define this if your system has /dev/urandom(4).
272HASSTRERROR Define this if you have the libc strerror(3) function (which
273 should be declared in <errno.h>), and it should be used
274 instead of sys_errlist.
275NEEDGETOPT Define this if you need a reimplementation of getopt(3).
276 On some systems, getopt does very odd things if called
277 to scan the arguments twice. This flag will ask sendmail
278 to compile in a local version of getopt that works
279 properly.
280NEEDSTRTOL Define this if your standard C library does not define
281 strtol(3). This will compile in a local version.
282NEEDVPRINTF Define this if your standard C library does not define
283 vprintf(3). Note that the resulting fake implementation
284 is not very elegant and may not even work on some
285 architectures.
286NEEDFSYNC Define this if your standard C library does not define
287 fsync(2). This will try to simulate the operation using
288 fcntl(2); if that is not available it does nothing, which
289 isn't great, but at least it compiles and runs.
290HASGETUSERSHELL Define this to 1 if you have getusershell(3) in your
291 standard C library. If this is not defined, or is defined
292 to be 0, sendmail will scan the /etc/shells file (no
293 NIS-style support, defaults to /bin/sh and /bin/csh if
294 that file does not exist) to get a list of unrestricted
295 user shells. This is used to determine whether users
296 are allowed to forward their mail to a program or a file.
297NEEDPUTENV Define this if your system needs am emulation of the
298 putenv(3) call. Define to 1 to implement it in terms
299 of setenv(3) or to 2 to do it in terms of primitives.
300NOFTRUNCATE Define this if you don't have the ftruncate(2) syscall.
301 If you don't have this system call, there is an unavoidable
302 race condition that occurs when creating alias databases.
303GIDSET_T The type of entries in a gidset passed as the second
304 argument to getgroups(2). Historically this has been an
305 int, so this is the default, but some systems (such as
306 IRIX) pass it as a gid_t, which is an unsigned short.
307 This will make a difference, so it is important to get
308 this right! However, it is only an issue if you have
309 group sets.
310SLEEP_T The type returned by the system sleep() function.
311 Defaults to "unsigned int". Don't worry about this
312 if you don't have compilation problems.
313ARBPTR_T The type of an arbitrary pointer -- defaults to "void *".
314 If you are an very old compiler you may need to define
315 this to be "char *".
316SOCKADDR_LEN_T The type used for the third parameter to accept(2),
317 getsockname(2), and getpeername(2), representing the
318 length of a struct sockaddr. Defaults to int.
319SOCKOPT_LEN_T The type used for the fifth parameter to getsockopt(2)
320 and setsockopt(2), representing the length of the option
321 buffer. Defaults to int.
322LA_TYPE The type of load average your kernel supports. These
323 can be one of:
324 LA_ZERO (1) -- it always returns the load average as
325 "zero" (and does so on all architectures).
326 LA_INT (2) to read /dev/kmem for the symbol avenrun and
327 interpret as a long integer.
328 LA_FLOAT (3) same, but interpret the result as a floating
329 point number.
330 LA_SHORT (6) to interpret as a short integer.
331 LA_SUBR (4) if you have the getloadavg(3) routine in your
332 system library.
333 LA_MACH (5) to use MACH-style load averages (calls
334 processor_set_info()),
335 LA_PROCSTR (7) to read /proc/loadavg and interpret it
336 as a string representing a floating-point
337 number (Linux-style).
338 LA_READKSYM (8) is an implementation suitable for some
339 versions of SVr4 that uses the MIOC_READKSYM ioctl
340 call to read /dev/kmem.
341 LA_DGUX (9) is a special implementation for DG/UX that uses
342 the dg_sys_info system call.
343 LA_HPUX (10) is an HP-UX specific version that uses the
344 pstat_getdynamic system call.
345 LA_IRIX6 (11) is an IRIX 6.x specific version that adapts
346 to 32 or 64 bit kernels; it is otherwise very similar
347 to LA_INT.
348 LA_KSTAT (12) uses the (Solaris-specific) kstat(3k)
349 implementation.
350 LA_DEVSHORT (13) reads a short from a system file (default:
351 /dev/table/avenrun) and scales it in the same manner
352 as LA_SHORT.
353 LA_INT, LA_SHORT, LA_FLOAT, and LA_READKSYM have several
354 other parameters that they try to divine: the name of your
355 kernel, the name of the variable in the kernel to examine,
356 the number of bits of precision in a fixed point load average,
357 and so forth. LA_DEVSHORT uses _PATH_AVENRUN to find the
358 device to be read to find the load average.
359 In desperation, use LA_ZERO. The actual code is in
360 conf.c -- it can be tweaked if you are brave.
361FSHIFT For LA_INT, LA_SHORT, and LA_READKSYM, this is the number
362 of bits of load average after the binary point -- i.e.,
363 the number of bits to shift right in order to scale the
364 integer to get the true integer load average. Defaults to 8.
365_PATH_UNIX The path to your kernel. Needed only for LA_INT, LA_SHORT,
366 and LA_FLOAT. Defaults to "/unix" on System V, "/vmunix"
367 everywhere else.
368LA_AVENRUN For LA_INT, LA_SHORT, and LA_FLOAT, the name of the kernel
369 variable that holds the load average. Defaults to "avenrun"
370 on System V, "_avenrun" everywhere else.
371SFS_TYPE Encodes how your kernel can locate the amount of free
372 space on a disk partition. This can be set to SFS_NONE
373 (0) if you have no way of getting this information,
374 SFS_USTAT (1) if you have the ustat(2) system call,
375 SFS_4ARGS (2) if you have a four-argument statfs(2)
376 system call (and the include file is <sys/statfs.h>),
377 SFS_VFS (3), SFS_MOUNT (4), SFS_STATFS (5) if you have
378 the two-argument statfs(2) system call with includes in
379 <sys/vfs.h>, <sys/mount.h>, or <sys/statfs.h> respectively,
380 or SFS_STATVFS (6) if you have the two-argument statvfs(2)
381 call. The default if nothing is defined is SFS_NONE.
382SFS_BAVAIL with SFS_4ARGS you can also set SFS_BAVAIL to the field name
383 in the statfs structure that holds the useful information;
384 this defaults to f_bavail.
385SPT_TYPE Encodes how your system can display what a process is doing
386 on a ps(1) command (SPT stands for Set Process Title). Can
387 be set to:
388 SPT_NONE (0) -- Don't try to set the process title at all.
389 SPT_REUSEARGV (1) -- Pad out your argv with the information;
390 this is the default if none specified.
391 SPT_BUILTIN (2) -- The system library has setproctitle.
392 SPT_PSTAT (3) -- Use the PSTAT_SETCMD option to pstat(2)
393 to set the process title; this is used by HP-UX.
394 SPT_PSSTRINGS (4) -- Use the magic PS_STRINGS pointer (4.4BSD).
395 SPT_SYSMIPS (5) -- Use sysmips() supported by NEWS-OS 6.
396 SPT_SCO (6) -- Write kernel u. area.
397 SPT_CHANGEARGV (7) -- Write pointers to our own strings into
398 the existing argv vector.
399SPT_PADCHAR Character used to pad the process title; if undefined,
400 the space character (0x20) is used. This is ignored if
401 SPT_TYPE != SPT_REUSEARGV
402ERRLIST_PREDEFINED
403 If set, assumes that some header file defines sys_errlist.
404 This may be needed if you get type conflicts on this
405 variable -- otherwise don't worry about it.
406WAITUNION The wait(2) routine takes a "union wait" argument instead
407 of an integer argument. This is for compatibility with
408 old versions of BSD.
409SCANF You can set this to extend the F command to accept a
410 scanf string -- this gives you a primitive parser for
411 class definitions -- BUT it can make you vulnerable to
412 core dumps if the target file is poorly formed.
413SYSLOG_BUFSIZE You can define this to be the size of the buffer that
414 syslog accepts. If it is not defined, it assumes a
415 1024-byte buffer. If the buffer is very small (under
416 256 bytes) the log message format changes -- each
417 e-mail message will log many more messages, since it
418 will log each piece of information as a separate line
419 in syslog.
420BROKEN_RES_SEARCH
421 On Ultrix (and maybe other systems?) if you use the
422 res_search routine with an unknown host name, it returns
423 -1 but sets h_errno to 0 instead of HOST_NOT_FOUND. If
424 you set this, sendmail considers 0 to be the same as
425 HOST_NOT_FOUND.
426NAMELISTMASK If defined, values returned by nlist(3) are masked
427 against this value before use -- a common value is
428 0x7fffffff to strip off the top bit.
429BSD4_4_SOCKADDR If defined, socket addresses have an sa_len field that
430 defines the length of this address.
431SAFENFSPATHCONF Set this to 1 if and only if you have verified that a
432 pathconf(2) call with _PC_CHOWN_RESTRICTED argument on an
433 NFS filesystem where the underlying system allows users to
434 give away files to other users returns <= 0. Be sure you
435 try both on NFS V2 and V3. Some systems assume that their
436 local policy apply to NFS servers -- this is a bad
437 assumption! The test/t_pathconf.c program will try this
438 for you -- you have to run it in a directory that is
439 mounted from a server that allows file giveaway.
440SIOCGIFCONF_IS_BROKEN
441 Set this if your system has an SIOCGIFCONF ioctl defined,
442 but it doesn't behave the same way as "most" systems (BSD,
443 Solaris, SunOS, HP-UX, etc.)
444SIOCGIFNUM_IS_BROKEN
445 Set this if your system has an SIOCGIFNUM ioctl defined,
446 but it doesn't behave the same way as "most" systems
447 (Solaris, HP-UX).
448NEED_PERCENTQ Set this if your system doesn't support the printf
449 format strings %lld or %llu. If this is set, %qd and
450 %qu are used instead.
451FAST_PID_RECYCLE
452 Set this if your system can reuse the same PID in the same
453 second.
454SO_REUSEADDR_IS_BROKEN
455 Set this if your system has a setsockopt() SO_REUSEADDR
456 flag but doesn't pay attention to it when trying to bind a
457 socket to a recently closed port.
458SNPRINTF_IS_BROKEN
459 Set this if your system has an snprintf() implementation
460 which does not NUL terminate the string being filled in.
461 Use test/t_snprintf.c to test your system.
462NEEDSGETIPNODE Set this if your system supports IPv6 but doesn't include
463 the getipnodeby{name,addr}() functions. Set automatically
464 for Linux's glibc.
465
466
467+-----------------------+
468| COMPILE-TIME FEATURES |
469+-----------------------+
470
471There are a bunch of features that you can decide to compile in, such
472as selecting various database packages and special protocol support.
473Several are assumed based on other compilation flags -- if you want to
474"un-assume" something, you probably need to edit conf.h. Compilation
475flags that add support for special features include:
476
477NDBM Include support for "new" DBM library for aliases and maps.
478 Normally defined in the Makefile.
479NEWDB Include support for Berkeley DB package (hash & btree)
480 for aliases and maps. Normally defined in the Makefile.
481 If the version of NEWDB you have is the old one that does
482 not include the "fd" call (this call was added in version
483 1.5 of the Berkeley DB code), you must upgrade to the
484 current version of Berkeley DB.
485NIS Define this to get NIS (YP) support for aliases and maps.
486 Normally defined in the Makefile.
487NISPLUS Define this to get NIS+ support for aliases and maps.
488 Normally defined in the Makefile.
489HESIOD Define this to get Hesiod support for aliases and maps.
490 Normally defined in the Makefile.
491NETINFO Define this to get NeXT NetInfo support for aliases and maps.
492 Normally defined in the Makefile.
493LDAPMAP Define this to get LDAP support for maps.
494PH_MAP Define this to get PH support for maps.
495MAP_NSD Define this to get nsd support for maps.
496USERDB Define this to 1 to include support for the User Information
497 Database. Implied by NEWDB or HESIOD. You can use
498 -DUSERDB=0 to explicitly turn it off.
499IDENTPROTO Define this as 1 to get IDENT (RFC 1413) protocol support.
500 This is assumed unless you are running on Ultrix or
501 HP-UX, both of which have a problem in the UDP
502 implementation. You can define it to be 0 to explicitly
503 turn off IDENT protocol support. If defined off, the code
504 is actually still compiled in, but it defaults off; you
505 can turn it on by setting the IDENT timeout in the
506 configuration file.
507IP_SRCROUTE Define this to 1 to get IP source routing information
508 displayed in the Received: header. This is assumed on
509 most systems, but some (e.g., Ultrix) apparently have a
510 broken version of getsockopt that doesn't properly
511 support the IP_OPTIONS call. You probably want this if
512 your OS can cope with it. Symptoms of failure will be that
513 it won't compile properly (that is, no support for fetching
514 IP_OPTIONs), or it compiles but source-routed TCP connections
515 either refuse to open or open and hang for no apparent reason.
516 Ultrix and AIX3 are known to fail this way.
517LOG Set this to get syslog(3) support. Defined by default
518 in conf.h. You want this if at all possible.
519NETINET Set this to get TCP/IP support. Defined by default
520 in conf.h. You probably want this.
521NETINET6 Set this to get IPv6 support. Other configuration may
522 be needed in conf.h for your particular operating system.
523 Also, DaemonPortOptions must be set appropriately for
524 sendmail to accept IPv6 connections.
525NETISO Define this to get ISO networking support.
526NETUNIX Define this to get Unix domain networking support. Defined
527 by default. A few bizarre systems (SCO, ISC, Altos) don't
528 support this networking domain.
529NETNS Define this to get NS networking support.
530NETX25 Define this to get X.25 networking support.
531SMTP Define this to get the SMTP code. Implied by NETINET
532 or NETISO.
533NAMED_BIND If non-zero, include DNS (name daemon) support, including
534 MX support. The specs say you must use this if you run
535 SMTP. You don't have to be running a name server daemon
536 on your machine to need this -- any use of the DNS resolver,
537 including remote access to another machine, requires this
538 option. Defined by default in conf.h. Define it to zero
539 ONLY on machines that do not use DNS in any way.
540QUEUE Define this to get queueing code. Implied by NETINET
541 or NETISO; required by SMTP. This gives you other good
542 stuff -- it should be on.
543DAEMON Define this to get general network support. Implied by
544 NETINET or NETISO. Defined by default in conf.h. You
545 almost certainly want it on.
546MATCHGECOS Permit fuzzy matching of user names against the full
547 name (GECOS) field in the /etc/passwd file. This should
548 probably be on, since you can disable it from the config
549 file if you want to. Defined by default in conf.h.
550MIME8TO7 If non-zero, include 8 to 7 bit MIME conversions. This
551 also controls advertisement of 8BITMIME in the ESMTP
552 startup dialogue.
553MIME7TO8 If non-zero, include 7 to 8 bit MIME conversions.
554HES_GETMAILHOST Define this to 1 if you are using Hesiod with the
555 hes_getmailhost() routine. This is included with the MIT
556 Hesiod distribution, but not with the DEC Hesiod distribution.
557XDEBUG Do additional internal checking. These don't cost too
558 much; you might as well leave this on.
559TCPWRAPPERS Turns on support for the TCP wrappers library (-lwrap).
560 See below for further information.
561SECUREWARE Enable calls to the SecureWare luid enabling/changing routines.
562 SecureWare is a C2 security package added to several UNIX's
563 (notably ConvexOS) to get a C2 Secure system. This
564 option causes mail delivery to be done with the luid of the
565 recipient.
566SHARE_V1 Support for the fair share scheduler, version 1. Setting to
567 1 causes final delivery to be done using the recipients
568 resource limitations. So far as I know, this is only
569 supported on ConvexOS.
570SASL Enables SMTP AUTH (RFC 2554). This requires the Cyrus SASL
571 library (ftp://ftp.andrew.cmu.edu/pub/cyrus-mail/). Please
572 install at least version 1.5.13. See below for further
573 information: SASL COMPILATION AND CONFIGURATION. If your
574 SASL library is older than 1.5.10, you have to set this
575 to its version number using a simple conversion: a.b.c
576 -> c + b*100 + a*10000, e.g. for 1.5.9 define SASL=10509.
577 Note: Using an older version than 1.5.5 of Cyrus SASL is
578 not supported. Starting with version 1.5.10, setting SASL=1
579 is sufficient. Any value other than 1 (or 0) will be
580 compared with the actual version found and if there is a
581 mismatch, compilation will fail.
582EGD Define this if your system has EGD installed, see
583 http://www.lothar.com/tech/crypto/ . It should be used to
584 seed the PRNG for STARTTLS if HASURANDOMDEV is not defined.
585STARTTLS Enables SMTP STARTTLS (RFC 2487). This requires OpenSSL
586 (http://www.OpenSSL.org/) and sfio (see below).
587 Use OpenSSL 0.9.5a or later (if compatible with this
588 version), do not use 0.9.3.
589 See STARTTLS COMPILATION AND CONFIGURATION for further
590 information.
591TLS_NO_RSA Turn off support for RSA algorithms in STARTTLS.
592SFIO Uses sfio instead of stdio. sfio is available from AT&T
593 (http://www.research.att.com/sw/tools/sfio/). If this
594 compile flag is set, confSTDIO_TYPE must be set to portable.
595 This compile flag is necessary for STARTTLS; it also
596 enables the security layer of SASL. The sfio include file
597 stdio.h must be installed in a subdirectory called sfio,
598 i.e., if you install sfio in /usr/local, stdio.h should
599 be in /usr/local/include/sfio, and libsfio.a should be in
600 /usr/local/lib. Notice: read the sfio section in
601 OPERATING SYSTEM AND COMPILE QUIRKS.
602
603
604Generic notice: If you enable a compile time option that needs
605libraries or include files that don't come with sendmail or are
606installed in a location that your C compiler doesn't use by default
607you should set confINCDIRS and confLIBDIRS as explained in the
608first section: BUILDING SENDMAIL.
609
610
611+---------------------+
612| DNS/RESOLVER ISSUES |
613+---------------------+
614
615Many systems have old versions of the resolver library. At a minimum,
616you should be running BIND 4.8.3; older versions may compile, but they
617have known bugs that should give you pause.
618
619Common problems in old versions include "undefined" errors for
620dn_skipname.
621
622Some people have had a problem with BIND 4.9; it uses some routines
623that it expects to be externally defined such as strerror(). It may
624help to link with "-l44bsd" to solve this problem. This has apparently
625been fixed in later versions of BIND, starting around 4.9.3. In other
626words, if you use 4.9.0 through 4.9.2, you need -l44bsd; for earlier or
627later versions, you do not.
628
629!PLEASE! be sure to link with the same version of the resolver as
630the header files you used -- some people have used the 4.9 headers
631and linked with BIND 4.8 or vice versa, and it doesn't work.
632Unfortunately, it doesn't fail in an obvious way -- things just
633subtly don't work.
634
635WILDCARD MX RECORDS ARE A BAD IDEA! The only situation in which they
636work reliably is if you have two versions of DNS, one in the real world
637which has a wildcard pointing to your firewall, and a completely
638different version of the database internally that does not include
639wildcard MX records that match your domain. ANYTHING ELSE WILL GIVE
640YOU HEADACHES!
641
642When attempting to canonify a hostname, some broken name servers will
643return SERVFAIL (a temporary failure) on T_AAAA (IPv6) lookups. If you
644want to excuse this behavior, compile sendmail with
645-D_FFR_WORKAROUND_BROKEN_NAMESERVERS and add WorkAroundBrokenAAAA to your
646ResolverOptions setting. However, instead, we recommend catching the
647problem and reporting it to the name server administrator so we can rid the
648world of broken name servers.
649
650
651+----------------------------------------+
652| STARTTLS COMPILATION AND CONFIGURATION |
653+----------------------------------------+
654
655Please read the docs accompanying the OpenSSL library and sfio.
656You have to compile and install both libraries before you can compile
657sendmail. See devtools/README how to set the correct compile time
658parameters; you should at least set the following variables:
659
660define(`confSTDIO_TYPE', `portable')
661APPENDDEF(`conf_sendmail_ENVDEF', `-DSFIO')
662APPENDDEF(`conf_sendmail_LIBS', `-lsfio')
663APPENDDEF(`conf_sendmail_ENVDEF', `-DSTARTTLS')
664APPENDDEF(`conf_sendmail_LIBS', `-lssl -lcrypto')
665
666If you have installed the OpenSSL libraries and include files in
667a location that your C compiler doesn't use by default you should
668set confINCDIRS and confLIBDIRS as explained in the first section:
669BUILDING SENDMAIL.
670
671Configuration information can be found in doc/op/op.me (required
672certificates) and cf/README (how to tell sendmail about certificates).
673
674To perform an initial test, connect to your sendmail daemon
675(telnet localhost 25) and issue a EHLO localhost and see whether
676250-STARTTLS
677is in the response. If it isn't, run the daemon with
678-O LogLevel=14
679and try again. Then take a look at the logfile and see whether
680there are any problems listed about permissions (unsafe files)
681or the validity of X.509 certificates.
682
683Note: sfio must be used in all libraries with which sendmail exchanges
684file pointers. An example is PH map support. This does not apply to the
685usual libraries, e.g., OpenSSL, Berkeley DB, Cyrus SASL.
686
687Further information can be found via:
688http://www.sendmail.org/tips/
689
690
691+------------------------------------+
692| SASL COMPILATION AND CONFIGURATION |
693+------------------------------------+
694
695Please read the docs accompanying the library (INSTALL and README).
696If you use Berkeley DB for Cyrus SASL then you must compile sendmail
697with the same version of Berkeley DB. See devtools/README how to
698set the correct compile time parameters; you should at least set
699the following variables:
700
701APPENDDEF(`conf_sendmail_ENVDEF', `-DSASL')
702APPENDDEF(`conf_sendmail_LIBS', `-lsasl')
703
704If you have installed the Cyrus SASL library and include files in
705a location that your C compiler doesn't use by default you should
706set confINCDIRS and confLIBDIRS as explained in the first section:
707BUILDING SENDMAIL.
708
709You have to select and install authentication mechanisms and tell
710sendmail where to find the sasl library and the include files (see
711devtools/README for the parameters to set). Setup the required
712users and passwords as explained in the SASL documentation. See
713also cf/README for authentication related options (esp. DefaultAuthInfo
714if you want authentication between MTAs).
715
716To perform an initial test, connect to your sendmail daemon
717(telnet localhost 25) and issue a EHLO localhost and see whether
718250-AUTH ....
719is in the response. If it isn't, run the daemon with
720-O LogLevel=14
721and try again. Then take a look at the logfile and see whether
722there are any security related problems listed (unsafe files).
723
724Further information can be found via:
725http://www.sendmail.org/tips/
726
727
728+-------------------------------------+
729| OPERATING SYSTEM AND COMPILE QUIRKS |
730+-------------------------------------+
731
732GCC problems
733 *****************************************************************
734 ** IMPORTANT: DO NOT USE OPTIMIZATION (``-O'') IF YOU ARE **
735 ** RUNNING GCC 2.4.x or 2.5.x. THERE IS A BUG IN THE GCC **
736 ** OPTIMIZER THAT CAUSES SENDMAIL COMPILES TO FAIL MISERABLY. **
737 *****************************************************************
738
739 Jim Wilson of Cygnus believes he has found the problem -- it will
740 probably be fixed in GCC 2.5.6 -- but until this is verified, be
741 very suspicious of gcc -O. This problem is reported to have been
742 fixed in gcc 2.6.
743
744 A bug in gcc 2.5.5 caused problems compiling sendmail 8.6.5 with
745 optimization on a Sparc. If you are using gcc 2.5.5, youi should
746 upgrade to the latest version of gcc.
747
748 Apparently GCC 2.7.0 on the Pentium processor has optimization
749 problems. I recommend against using -O on that architecture. This
750 has been seen on FreeBSD 2.0.5 RELEASE.
751
752 Solaris 2.X users should use version 2.7.2.3 over 2.7.2.
753
754 We have been told there are problems with gcc 2.8.0. If you are
755 using this version, you should upgrade to 2.8.1 or later.
756
757GDBM GDBM does not work with sendmail 8.8 because the additional
758 security checks and file locking cause problems. Unfortunately,
759 gdbm does not provide a compile flag in its version of ndbm.h so
760 the code can adapt. Until the GDBM authors can fix these problems,
761 GDBM will not be supported. Please use Berkeley DB instead.
762
763Configuration file location
764 Up to 8.6, sendmail tried to find the sendmail.cf file in the same
765 place as the vendors had put it, even when this was obviously
766 stupid. As of 8.7, sendmail ALWAYS looks for /etc/sendmail.cf.
767 Beginning with 8.10, sendmail uses /etc/mail/sendmail.cf.
768 You can get sendmail to use the stupid vendor .cf location by
769 adding -DUSE_VENDOR_CF_PATH during compilation, but this may break
770 support programs and scripts that need to find sendmail.cf. You
771 are STRONGLY urged to use symbolic links if you want to use the
772 vendor location rather than changing the location in the sendmail
773 binary.
774
775 NETINFO systems use NETINFO to determine the location of
776 sendmail.cf. The full path to sendmail.cf is stored as the value of
777 the "sendmail.cf" property in the "/locations/sendmail"
778 subdirectory of NETINFO. Set the value of this property to
779 "/etc/mail/sendmail.cf" (without the quotes) to use this new
780 default location for Sendmail 8.10.0 and higher.
781
782ControlSocket permissions
783 Paraphrased from BIND 8.2.1's README:
784
785 Solaris and other pre-4.4BSD kernels do not respect ownership or
786 protections on UNIX-domain sockets. The short term fix for this is to
787 override the default path and put such control sockets into root-
788 owned directories which do not permit non-root to r/w/x through them.
789 The long term fix is for all kernels to upgrade to 4.4BSD semantics.
790
791SunOS 4.x (Solaris 1.x)
792 You may have to use -lresolv on SunOS. However, beware that
793 this links in a new version of gethostbyname that does not
794 understand NIS, so you must have all of your hosts in DNS.
795
796 Some people have reported problems with the SunOS version of
797 -lresolv and/or in.named, and suggest that you get a newer
798 version. The symptoms are delays when you connect to the
799 SMTP server on a SunOS machine or having your domain added to
800 addresses inappropriately. There is a version of BIND
801 version 4.9 on gatekeeper.DEC.COM in pub/BSD/bind/4.9.
802
803 There is substantial disagreement about whether you can make
804 this work with resolv+, which allows you to specify a search-path
805 of services. Some people report that it works fine, others
806 claim it doesn't work at all (including causing sendmail to
807 drop core when it tries to do multiple resolv+ lookups for a
808 single job). I haven't tried resolv+, as we use DNS exclusively.
809
810 Should you want to try resolv+, it is on ftp.uu.net in
811 /networking/ip/dns.
812
813 Apparently getservbyname() can fail under moderate to high
814 load under some circumstances. This will exhibit itself as
815 the message ``554 makeconnection: service "smtp" unknown''.
816 The problem has been traced to one or more blank lines in
817 /etc/services on the NIS server machine. Delete these
818 and it should work. This info is thanks to Brian Bartholomew
819 <bb@math.ufl.edu> of I-Kinetics, Inc.
820
821 NOTE: The SunOS 4.X linker uses library paths specified during
822 compilation using -L for run-time shared library searches.
823 Therefore, it is vital that relative and unsafe directory paths not
824 be used when compiling sendmail.
825
826SunOS 4.0.2 (Sun 386i)
827 Date: Fri, 25 Aug 1995 11:13:58 +0200 (MET DST)
828 From: teus@oce.nl
829
830 Sendmail 8.7.Beta.12 compiles and runs nearly out of the box with the
831 following changes:
832 * Don't use /usr/5bin in your PATH, but make /usr/5bin/uname
833 available as "uname" command.
834 * Use the defines "-DBSD4_3 -DNAMED_BIND=0" in
835 devtools/OS/SunOS.4.0, which is selected via the "uname" command.
836 I recommend to make available the db-library on the system first
837 (and change the Makefile to use this library).
838 Note that the sendmail.cf and aliases files are found in /etc.
839
840SunOS 4.1.3, 4.1.3_U1
841 Sendmail causes crashes on SunOS 4.1.3 and 4.1.3_U1. According
842 to Sun bug number 1077939:
843
844 If an application does a getsockopt() on a SOCK_STREAM (TCP) socket
845 after the other side of the connection has sent a TCP RESET for
846 the stream, the kernel gets a Bus Trap in the tcp_ctloutput() or
847 ip_ctloutput() routine.
848
849 For 4.1.3, this is fixed in patch 100584-08, available on the
850 Sunsolve 2.7.1 or later CDs. For 4.1.3_U1, this was fixed in patch
851 101790-01 (SunOS 4.1.3_U1: TCP socket and reset problems), later
852 obsoleted by patch 102010-05.
853
854 Sun patch 100584-08 is not currently publicly available on their
855 ftp site but a user has reported it can be found at other sites
856 using a web search engine.
857
858Solaris 2.x (SunOS 5.x)
859 To compile for Solaris, the Makefile built by Build must
860 include a SOLARIS definition which reflects the Solaris version
861 (i.e. -DSOLARIS=20400 for 2.4 or -DSOLARIS=20501 for 2.5.1).
862 If you are using gcc, make sure -I/usr/include is not used (or
863 it might complain about TopFrame). If you are using Sun's cc,
864 make sure /opt/SUNWspro/bin/cc is used instead of /usr/ucb/cc
865 (or it might complain about tm_zone).
866
867 The Solaris "syslog" function is apparently limited to something
868 about 90 characters because of a kernel limitation. If you have
869 source code, you can probably up this number. You can get patches
870 that fix this problem: the patch ids are:
871
872 Solaris 2.1 100834
873 Solaris 2.2 100999
874 Solaris 2.3 101318
875
876 Be sure you have the appropriate patch installed or you won't
877 see system logging.
878
879Solaris 2.4 (SunOS 5.4)
880 If you include /usr/lib at the end of your LD_LIBRARY_PATH you run
881 the risk of getting the wrong libraries under some circumstances.
882 This is because of a new feature in Solaris 2.4, described by
883 Rod.Evans@Eng.Sun.COM:
884
885 >> Prior to SunOS 5.4, any LD_LIBRARY_PATH setting was ignored by the
886 >> runtime linker if the application was setxid (secure), thus your
887 >> applications search path would be:
888 >>
889 >> /usr/local/lib LD_LIBRARY_PATH component - IGNORED
890 >> /usr/lib LD_LIBRARY_PATH component - IGNORED
891 >> /usr/local/lib RPATH - honored
892 >> /usr/lib RPATH - honored
893 >>
894 >> the effect is that path 3 would be the first used, and this would
895 >> satisfy your resolv.so lookup.
896 >>
897 >> In SunOS 5.4 we made the LD_LIBRARY_PATH a little more flexible.
898 >> People who developed setxid applications wanted to be able to alter
899 >> the library search path to some degree to allow for their own
900 >> testing and debugging mechanisms. It was decided that the only
901 >> secure way to do this was to allow a `trusted' path to be used in
902 >> LD_LIBRARY_PATH. The only trusted directory we presently define
903 >> is /usr/lib. Thus a setuid root developer could play with some
904 >> alternative shared object implementations and place them in
905 >> /usr/lib (being root we assume they'ed have access to write in this
906 >> directory). This change was made as part of 1155380 - after a
907 >> *huge* amount of discussion regarding the security aspect of things.
908 >>
909 >> So, in SunOS 5.4 your applications search path would be:
910 >>
911 >> /usr/local/lib from LD_LIBRARY_PATH - IGNORED (untrustworthy)
912 >> /usr/lib from LD_LIBRARY_PATH - honored (trustworthy)
913 >> /usr/local/lib from RPATH - honored
914 >> /usr/lib from RPATH - honored
915 >>
916 >> here, path 2 would be the first used.
917
918Solaris 2.5.1 (SunOS 5.5.1) and 2.6 (SunOS 5.6)
919 Apparently Solaris 2.5.1 patch 103663-01 installs a new
920 /usr/include/resolv.h file that defines the __P macro without
921 checking to see if it is already defined. This new resolv.h is also
922 included in the Solaris 2.6 distribution. This causes compile
923 warnings such as:
924
925 In file included from daemon.c:51:
926 /usr/include/resolv.h:208: warning: `__P' redefined
927 cdefs.h:58: warning: this is the location of the previous definition
928
929 These warnings can be safely ignored or you can create a resolv.h
930 file in the obj.SunOS.5.5.1.* or obj.SunOS.5.6.* directory that reads:
931
932 #undef __P
933 #include "/usr/include/resolv.h"
934
935 Sun is aware of the problem (Sun bug ID 4081053) and it will be fixed
936 in Solaris 2.7.
937
938Solaris 7 (SunOS 5.7)
939 Solaris 7 includes LDAP libraries but the implementation was
940 lacking a few things. The following settings can be placed in
941 devtools/Site/site.SunOS.5.7.m4 if you plan on using those
942 libraries.
943
944 APPENDDEF(`confMAPDEF', `-DLDAPMAP')
945 APPENDDEF(`confENVDEF', `-DLDAP_VERSION_MAX=3')
946 APPENDDEF(`confLIBS', `-lldap')
947
948 Also, Sun's patch 107555 is needed to prevent a crash in the call
949 to ldap_set_option for LDAP_OPT_REFERRALS in ldapmap_setopts if
950 LDAP support is compiled in sendmail.
951
952Solaris
953 If you are using dns for hostname resolution on Solaris, make sure
954 that the 'dns' entry is last on the hosts line in
955 '/etc/nsswitch.conf'. For example, use:
956
957 hosts: nisplus files dns
958
959 Do not use:
960
961 host: nisplus dns [NOTFOUND=return] files
962
963 Note that 'nisplus' above is an illustration. The same comment
964 applies no matter what naming services you are using. If you have
965 anything other than dns last, even after "[NOTFOUND=return]",
966 sendmail may not be able to determine whether an error was
967 temporary or permanent. The error returned by the solaris
968 gethostbyname() is the error for the last lookup used, and other
969 naming services do not have the same concept of temporary failure.
970
971Ultrix
972 By default, the IDENT protocol is turned off on Ultrix. If you
973 are running Ultrix 4.4 or later, or if you have included patch
974 CXO-8919 for Ultrix 4.2 or 4.3 to fix the TCP problem, you can turn
975 IDENT on in the configuration file by setting the "ident" timeout.
976
977 The Ultrix 4.5 Y2K patch (ULTV45-022-1) has changed the resolver
978 included in libc.a. Unfortunately, the __RES symbol hasn't changed
979 and therefore, sendmail can no longer automatically detect the
980 newer version. If you get a compiler error:
981
982 /lib/libc.a(gethostent.o): local_hostname_length: multiply defined
983
984 Then rebuild with this in devtools/Site/site.ULTRIX.m4:
985
986 APPENDDEF(`conf_sendmail_ENVDEF', `-DNEEDLOCAL_HOSTNAME_LENGTH=0')
987
988Digital UNIX (formerly DEC OSF/1)
989 If you are compiling on OSF/1 (DEC Alpha), you must use
990 -L/usr/shlib (otherwise it core dumps on startup). You may also
991 need -mld to get the nlist() function, although some versions
992 apparently don't need this.
993
994 Also, the enclosed makefile removed /usr/sbin/smtpd; if you need
995 it, just create the link to the sendmail binary.
996
997 On DEC OSF/1 3.2 or earlier, the MatchGECOS option doesn't work
998 properly due to a bug in the getpw* routines. If you want to use
999 this, use -DDEC_OSF_BROKEN_GETPWENT=1. The problem is fixed in 3.2C.
1000
1001 Digital's mail delivery agent, /bin/mail (aka /bin/binmail), will
1002 only preserve the envelope sender in the "From " header if
1003 DefaultUserID is set to daemon. Setting this to mailnull will
1004 cause all mail to have the header "From mailnull ...". To use
1005 a different DefaultUserID, you will need to use a different mail
1006 delivery agent (such as mail.local found in the sendmail
1007 distribution).
1008
1009 On Digital UNIX 4.0 and later, Berkeley DB 1.85 is included with the
1010 operating system and already has the ndbm.o module removed. However,
1011 Digital has modified the original Berkeley DB db.h include file.
1012 This results in the following warning while compiling map.c and udb.c:
1013
1014 cc: Warning: /usr/include/db.h, line 74: The redefinition of the macro
1015 "__signed" conflicts with a current definition because the replacement
1016 lists differ. The redefinition is now in effect.
1017 #define __signed signed
1018 ------------------------^
1019
1020 This warning can be ignored.
1021
1022 Digital UNIX's linker checks /usr/ccs/lib/ before /usr/lib/.
1023 If you have installed a new version of BIND in /usr/include
1024 and /usr/lib, you will experience difficulties as Digital ships
1025 libresolv.a in /usr/ccs/lib/ as well. Be sure to replace both
1026 copies of libresolv.a.
1027
1028IRIX
1029 The header files on SGI IRIX are completely prototyped, and as
1030 a result you can sometimes get some warning messages during
1031 compilation. These can be ignored. There are two errors in
1032 deliver only if you are using gcc, both of the form ``warning:
1033 passing arg N of `execve' from incompatible pointer type''.
1034 Also, if you compile with -DNIS, you will get a complaint
1035 about a declaration of struct dom_binding in a prototype
1036 when compiling map.c; this is not important because the
1037 function being prototyped is not used in that file.
1038
1039 In order to compile sendmail you will have had to install
1040 the developers' option in order to get the necessary include
1041 files.
1042
1043 If you compile with -lmalloc (the fast memory allocator), you may
1044 get warning messages such as the following:
1045
1046 ld32: WARNING 85: definition of _calloc in /usr/lib32/libmalloc.so
1047 preempts that definition in /usr/lib32/mips3/libc.so.
1048 ld32: WARNING 85: definition of _malloc in /usr/lib32/libmalloc.so
1049 preempts that definition in /usr/lib32/mips3/libc.so.
1050 ld32: WARNING 85: definition of _realloc in /usr/lib32/libmalloc.so
1051 preempts that definition in /usr/lib32/mips3/libc.so.
1052 ld32: WARNING 85: definition of _free in /usr/lib32/libmalloc.so
1053 preempts that definition in /usr/lib32/mips3/libc.so.
1054 ld32: WARNING 85: definition of _cfree in /usr/lib32/libmalloc.so
1055 preempts that definition in /usr/lib32/mips3/libc.so.
1056
1057 These are unavoidable and innocuous -- just ignore them.
1058
1059 According to Dave Sill <de5@ornl.gov>, there is a version of the
1060 Berkeley DB library patched to run on Irix 6.2 available from
1061 http://reality.sgi.com/ariel/freeware/#db .
1062
1063IRIX 6.x
1064 If you are using XFS filesystem, avoid using the -32 ABI switch to
1065 the cc compiler if possible.
1066
1067 Broken inet_aton and inet_ntoa on IRIX using gcc: There's
1068 a problem with gcc on IRIX, i.e., gcc can't pass structs
1069 less than 16 bits long unless they are 8 bits; IRIX 6.2 has
1070 some other sized structs. See
1071 http://www.bitmechanic.com/mail-archives/mysql/current/0418.html
1072
1073IRIX 6.4
1074 The IRIX 6.5.4 version of /bin/m4 does not work properly with
1075 sendmail. Either install fw_m4.sw.m4 off the Freeware_May99 CD and
1076 use /usr/freeware/bin/m4 or install and use GNU m4.
1077
1078NeXT or NEXTSTEP
1079 NEXTSTEP 3.3 and earlier ship with the old DBM library. Also,
1080 Berkeley DB does not currently run on NEXTSTEP.
1081
1082 If you are compiling on NEXTSTEP, you will have to create an
1083 empty file "unistd.h" and create a file "dirent.h" containing:
1084
1085 #include <sys/dir.h>
1086 #define dirent direct
1087
1088 (devtools/OS/NeXT should try to do both of these for you.)
1089
1090 Apparently, there is a bug in getservbyname on Nextstep 3.0
1091 that causes it to fail under some circumstances with the
1092 message "SYSERR: service "smtp" unknown" logged. You should
1093 be able to work around this by including the line:
1094
1095 OOPort=25
1096
1097 in your .cf file.
1098
1099BSDI (BSD/386) 1.0, NetBSD 0.9, FreeBSD 1.0
1100 The "m4" from BSDI won't handle the config files properly.
1101 I haven't had a chance to test this myself.
1102
1103 The M4 shipped in FreeBSD and NetBSD 0.9 don't handle the config
1104 files properly. One must use either GNU m4 1.1 or the PD-M4
1105 recently posted in comp.os.386bsd.bugs (and maybe others).
1106 NetBSD-current includes the PD-M4 (as stated in the NetBSD file
1107 CHANGES).
1108
1109 FreeBSD 1.0 RELEASE has uname(2) now. Use -DUSEUNAME in order to
1110 use it (look into devtools/OS/FreeBSD). NetBSD-current may have
1111 it too but it has not been verified.
1112
1113 The latest version of Berkeley DB uses a different naming
1114 scheme than the version that is supplied with your release. This
1115 means you will be able to use the current version of Berkeley DB
1116 with sendmail as long you use the new db.h when compiling
1117 sendmail and link it against the new libdb.a or libdb.so. You
1118 should probably keep the original db.h in /usr/include and the
1119 new db.h in /usr/local/include.
1120
11214.3BSD
1122 If you are running a "virgin" version of 4.3BSD, you'll have
1123 a very old resolver and be missing some header files. The
1124 header files are simple -- create empty versions and everything
1125 will work fine. For the resolver you should really port a new
1126 version (4.8.3 or later) of the resolver; 4.9 is available on
1127 gatekeeper.DEC.COM in pub/BSD/bind/4.9. If you are really
1128 determined to continue to use your old, buggy version (or as
1129 a shortcut to get sendmail working -- I'm sure you have the
1130 best intentions to port a modern version of BIND), you can
1131 copy ../contrib/oldbind.compat.c into sendmail and add the
1132 following to devtools/Site/site.config.m4:
1133
1134 APPENDDEF(`confOBJADD', `oldbind.compat.o')
1135
1136A/UX
1137 Date: Tue, 12 Oct 1993 18:28:28 -0400 (EDT)
1138 From: "Eric C. Hagberg" <hagberg@med.cornell.edu>
1139 Subject: Fix for A/UX ndbm
1140
1141 I guess this isn't really a sendmail bug, however, it is something
1142 that A/UX users should be aware of when compiling sendmail 8.6.
1143
1144 Apparently, the calls that sendmail is using to the ndbm routines
1145 in A/UX 3.0.x contain calls to "broken" routines, in that the
1146 aliases database will break when it gets "just a little big"
1147 (sorry I don't have exact numbers here, but it broke somewhere
1148 around 20-25 aliases for me.), making all aliases non-functional
1149 after exceeding this point.
1150
1151 What I did was to get the gnu-dbm-1.6 package, compile it, and
1152 then re-compile sendmail with "-lgdbm", "-DNDBM", and using the
1153 ndbm.h header file that comes with the gnu-package. This makes
1154 things behave properly.
1155 [NOTE: see comment above about GDBM]
1156
1157 I suppose porting the New Berkeley DB package is another route,
1158 however, I made a quick attempt at it, and found it difficult
1159 (not easy at least); the gnu-dbm package "configured" and
1160 compiled easily.
1161
1162 [NOTE: Berkeley DB version 2.X runs on A/UX and can be used for
1163 database maps.]
1164
1165SCO Unix
1166 From: Thomas Essebier <tom@stallion.oz.au>
1167 Organisation: Stallion Technologies Pty Ltd.
1168
1169 It will probably help those who are trying to configure sendmail 8.6.9
1170 to know that if they are on SCO, they had better set
1171 OI-dnsrch
1172 or they will core dump as soon as they try to use the resolver.
1173 ie. although SCO has _res.dnsrch defined, and is kinda BIND 4.8.3, it
1174 does not inititialise it, nor does it understand 'search' in
1175 /etc/named.boot.
1176 - sigh -
1177
1178 According to SCO, the m4 which ships with UnixWare 2.1.2 is broken.
1179 We recommend installing GNU m4 before attempting to build sendmail.
1180
1181DG/UX
1182 Doug Anderson <dlander@afterlife.ncsc.mil> has successfully run
1183 V8 on the DG/UX 5.4.2 and 5.4R3.x platforms under heavy usage.
1184 Originally, the DG /bin/mail program wasn't compatible with
1185 the V8 sendmail, since the DG /bin/mail requires the environment
1186 variable "_FORCE_MAIL_LOCAL_=yes" be set. Version 8.7 now includes
1187 this in the environment before invoking the local mailer. Some
1188 have used procmail to avoid this problem in the past. It works
1189 but some have experienced file locking problems with their DG/UX
1190 ports of procmail.
1191
1192Apollo DomainOS
1193 If you are compiling on Apollo, you will have to create an empty
1194 file "unistd.h" (for DomainOS 10.3 and earlier) and create a file
1195 "dirent.h" containing:
1196
1197 #include <sys/dir.h>
1198 #define dirent direct
1199
1200 (devtools/OS/DomainOS will attempt to do both of these for you.)
1201
1202HP-UX 8.00
1203 Date: Mon, 24 Jan 1994 13:25:45 +0200
1204 From: Kimmo Suominen <Kimmo.Suominen@lut.fi>
1205 Subject: 8.6.5 w/ HP-UX 8.00 on s300
1206
1207 Just compiled and fought with sendmail 8.6.5 on a HP9000/360 (ie. a
1208 series 300 machine) running HP-UX 8.00.
1209
1210 I was getting segmentation fault when delivering to a local user.
1211 With debugging I saw it was faulting when doing _free@libc... *sigh*
1212 It seems the new implementation of malloc on s300 is buggy as of 8.0,
1213 so I tried out the one in -lmalloc (malloc(3X)). With that it seems
1214 to work just dandy.
1215
1216 When linking, you will get the following error:
1217
1218 ld: multiply defined symbol _freespace in file /usr/lib/libmalloc.a
1219
1220 but you can just ignore it. You might want to add this info to the
1221 README file for the future...
1222
1223Linux
1224 Something broke between versions 0.99.13 and 0.99.14 of Linux:
1225 the flock() system call gives errors. If you are running .14,
1226 you must not use flock. You can do this with -DHASFLOCK=0.
1227
1228 Around the inclusion of bind-4.9.3 & Linux libc-4.6.20, the
1229 initialization of the _res structure changed. If /etc/hosts.conf
1230 was configured as "hosts, bind" the resolver code could return
1231 "Name server failure" errors. This is supposedly fixed in
1232 later versions of libc (>= 4.6.29?), and later versions of
1233 sendmail (> 8.6.10) try to work around the problem.
1234
1235 Some older versions (< 4.6.20?) of the libc/include files conflict
1236 with sendmail's version of cdefs.h. Deleting sendmail's version
1237 on those systems should be non-harmful, and new versions don't care.
1238
1239 Sendmail assumes that libc has snprintf, which has been true since
1240 libc 4.7.0. If you are running an older version, you will need to
1241 use -DHASSNPRINTF=0 in the Makefile. If may be able to use -lbsd
1242 (which includes snprintf) instead of turning this off on versions
1243 of libc between 4.4.4 and 4.7.0 (snprintf improves security, so
1244 you want to use this if at all possible).
1245
1246 NOTE ON LINUX & BIND: By default, the Makefile generated for Linux
1247 includes header files in /usr/local/include and libraries in
1248 /usr/local/lib. If you've installed BIND on your system, the header
1249 files typically end up in the search path and you need to add
1250 "-lresolv" to the LIBS line in your Makefile. Really old versions
1251 may need to include "-l44bsd" as well (particularly if the link phase
1252 complains about missing strcasecmp, strncasecmp or strpbrk).
1253 Complaints about an undefined reference to `__dn_skipname' in
1254 domain.o are a sure sign that you need to add -lresolv to LIBS.
1255 Newer versions of Linux are basically threaded BIND, so you may or
1256 may not see complaints if you accidentally mix BIND
1257 headers/libraries with virginal libc. If you have BIND headers in
1258 /usr/local/include (resolv.h, etc) you *should* be adding -lresolv
1259 to LIBS. Data structures may change and you'd be asking for a
1260 core dump.
1261
1262 A number of problems have been reported regarding the Linux 2.2.0
1263 kernel. So far, these problems have been tracked down to syslog()
1264 and DNS resolution. We believe the problem is with the poll()
1265 implementation in the Linux 2.2.0 kernel and poll()-aware versions
1266 of glib (at least up to 2.0.111).
1267
1268 Some pre-glibc distributions of Linux include a syslog.h that does
1269 not work properly with SFIO. You can fix this by adding
1270 "#include <syslog.h>" to the SFIO version of stdio.h as the very
1271 first line.
1272
1273glibc
1274 glibc 2.2.1 (and possibly other versions) changed the value of
1275 __RES in resolv.h but failed to actually provide the IPv6 API
1276 changes that the change implied. Therefore, compiling with
1277 -DNETINET6 fails.
1278
1279 Workarounds:
1280 1) Compile without -DNETINET6
1281 2) Build against a real BIND 8.2.2 include/lib tree
1282 3) Wait for glibc to fix it
1283
1284AIX 4.X
1285 The AIX 4.X linker uses library paths specified during compilation
1286 using -L for run-time shared library searches. Therefore, it is
1287 vital that relative and unsafe directory paths not be using when
1288 compiling sendmail. Because of this danger, by default, compiles
1289 on AIX use the -blibpath option to limit shared libraries to
1290 /usr/lib and /lib. If you need to allow more directories, such as
1291 /usr/local/lib, modify your devtools/Site/site.AIX.4.2.m4,
1292 site.AIX.4.3.m4, and/or site.AIX.4.x.m4 file(s) and set confLDOPTS
1293 approriately. For example:
1294
1295 define(`confLDOPTS', `-blibpath:/usr/lib:/lib:/usr/local/lib')
1296
1297 Be sure to only add (safe) system directories.
1298
1299 The AIX version of GNU ld also exhibits this problem. If you are
1300 using that version, instead of -blibpath, use its -rpath option.
1301 For example:
1302
1303 gcc -Wl,-rpath /usr/lib -Wl,-rpath /lib -Wl,-rpath /usr/local/lib
1304
1305AIX 4.3.3
1306 From: Valdis.Kletnieks@vt.edu
1307 Date: Sun, 02 Jul 2000 03:58:02 -0400
1308
1309 Under AIX 4.3.3, after applying bos.adt.include 4.3.3.12 to close the
1310 BIND 8.2.2 security holes, you can no longer build with -DNETINET6
1311 because they changed the value of __RES in resolv.h but failed to
1312 actually provide the API changes that the change implied.
1313
1314 Workarounds:
1315 1) Compile without -DNETINET6
1316 2) Build against a real BIND 8.2.2 include/lib tree
1317 3) Wait for IBM to fix it
1318
1319AIX 4.X
1320 The AIX m4 implements a different mechanism for ifdef which is
1321 inconsistent with other versions of m4. Therefore, it will not
1322 work properly with the sendmail Build architecture or m4
1323 configuration method. To work around this problem, please use
1324 GNU m4 from ftp://ftp.gnu.org/pub/gnu/.
1325
1326AIX 3.x
1327 This version of sendmail does not support MB, MG, and MR resource
1328 records, which are supported by AIX sendmail.
1329
1330 Several people have reported that the IBM-supplied named returns
1331 fairly random results -- the named should be replaced. It is not
1332 necessary to replace the resolver, which will simplify installation.
1333 A new BIND resolver can be found at http://www.isc.org/isc/.
1334
1335AIX 3.1.x
1336 The supplied load average code only works correctly for AIX 3.2.x.
1337 For 3.1, use -DLA_TYPE=LA_SUBR and get the latest ``monitor''
1338 package by Jussi Maki <jmaki@hut.fi> from ftp.funet.fi in the
1339 directory pub/unix/AIX/rs6000/monitor-1.12.tar.Z; use the loadavgd
1340 daemon, and the getloadavg subroutine supplied with that package.
1341 If you don't care about load average throttling, just turn off
1342 load average checking using -DLA_TYPE=LA_ZERO.
1343
1344AIX 2.2.1
1345 Date: Mon Dec 4 14:14:56 CST 1995
1346 From: Mark Whetzel <markw@antimatr.houston.tx.us>
1347 Subject: Porting sendmail 8.7.2 to AIX V2 on the RT.
1348
1349 This version of sendmail does not support MB, MG, and MR resource
1350 records, which are supported by AIX sendmail.
1351
1352 AIX V2 on the RT does not have 'paths.h'. Create a null
1353 file in the 'obj' directory to remove this compile error.
1354
1355 A patch file is needed to get the BSD 'db' library to compile
1356 for AIX/RT. I have sent the necessary updates to the author,
1357 but they may not be immediately available.
1358 [NOTE: Berkeley DB version 2.X runs on AIX/RT.]
1359
1360 The original AIX/RT resolver libraries are very old, and you
1361 should get the latest BIND to replace it. The 4.8.3 version
1362 has been tested, but 4.9.x is out and should work.
1363
1364 To make the load average code work correctly requires an
1365 external routine, as the kernel does not maintain system
1366 load averages, similar to AIX V3.1.x. A reverse port of the
1367 older 1.05 'monitor' load average daemon code written by
1368 Jussi Maki that will work on AIX V2 for the RT is available
1369 by E-mail to Mark Whetzel <markw@antimatr.houston.tx.us>.
1370 That code depends on an external daemon to collect system
1371 load information, and the external routine 'getloadavg',
1372 that will return that information. The 'LA_SUBR' define
1373 will handle this for AIX V2 on the RT.
1374
1375 Note: You will have to change devtools/OS/AIX.2 to correctly
1376 point to the locatons of the updated BIND source tree and
1377 the location of the 'newdb' tree and library location.
1378 You will also have to change devtools/OS/AIX.2 to know
1379 about the location of the 'getloadavg' routine if you use
1380 the LA_SUBR define.
1381
1382RISC/os
1383 RISC/os from MIPS is a merged AT&T/Berkeley system. When you
1384 compile on that platform you will get duplicate definitions
1385 on many files. You can ignore these.
1386
1387System V Release 4 Based Systems
1388 There is a single devtools OS that is intended for all SVR4-based
1389 systems (built from devtools/OS/SVR4). It defines __svr4__,
1390 which is predefined by some compilers. If your compiler already
1391 defines this compile variable, you can delete the definition from
1392 the generated Makefile or create a devtools/Site/site.config.m4
1393 file.
1394
1395 It's been tested on Dell Issue 2.2.
1396
1397DELL SVR4
1398 Date: Mon, 06 Dec 1993 10:42:29 EST
1399 From: "Kimmo Suominen" <kim@grendel.lut.fi>
1400 Message-ID: <2d0352f9.lento29@lento29.UUCP>
1401 To: eric@cs.berkeley.edu
1402 Cc: sendmail@cs.berkeley.edu
1403 Subject: Notes for DELL SVR4
1404
1405 Eric,
1406
1407 Here are some notes for compiling Sendmail 8.6.4 on DELL SVR4. I ran
1408 across these things when helping out some people who contacted me by
1409 e-mail.
1410
1411 1) Use gcc 2.4.5 (or later?). Dell distributes gcc 2.1 with their
1412 Issue 2.2 Unix. It is too old, and gives you problems with
1413 clock.c, because sigset_t won't get defined in <sys/signal.h>.
1414 This is due to a problematic protection rule in there, and is
1415 fixed with gcc 2.4.5.
1416
1417 2) If you don't use the new Berkeley DB (-DNEWDB), then you need
1418 to add "-lc -lucb" to the libraries to link with. This is because
1419 the -ldbm distributed by Dell needs the bcopy, bcmp and bzero
1420 functions. It is important that you specify both libraries in
1421 the given order to be sure you only get the BSTRING functions
1422 from the UCB library (and not the signal routines etc.).
1423
1424 3) Don't leave out "-lelf" even if compiling with "-lc -lucb".
1425 The UCB library also has another copy of the nlist routines,
1426 but we do want the ones from "-lelf".
1427
1428 If anyone needs a compiled gcc 2.4.5 and/or a ported DB library, they
1429 can use anonymous ftp to fetch them from lut.fi in the /kim directory.
1430 They are copies of what I use on grendel.lut.fi, and offering them
1431 does not imply that I would also support them. I have sent the DB
1432 port for SVR4 back to Keith Bostic for inclusion in the official
1433 distribution, but I haven't heard anything from him as of today.
1434
1435 - gcc-2.4.5-svr4.tar.gz (gcc 2.4.5 and the corresponding libg++)
1436 - db-1.72.tar.gz (with source, objects and a installed copy)
1437
1438 Cheers
1439 + Kim
1440 --
1441 * Kimmo.Suominen@lut.fi * SysVr4 enthusiast at GRENDEL.LUT.FI *
1442 * KIM@FINFILES.BITNET * Postmaster and Hostmaster at LUT.FI *
1443 * + 358 200 865 718 * Unix area moderator at NIC.FUNET.FI *
1444
1445ConvexOS 10.1 and below
1446 In order to use the name server, you must create the file
1447 /etc/use_nameserver. If this file does not exist, the call
1448 to res_init() will fail and you will have absolutely no
1449 access to DNS, including MX records.
1450
1451Amdahl UTS 2.1.5
1452 In order to get UTS to work, you will have to port BIND 4.9.
1453 The vendor's BIND is reported to be ``totally inadequate.''
1454 See sendmail/contrib/AmdahlUTS.patch for the patches necessary
1455 to get BIND 4.9 compiled for UTS.
1456
1457UnixWare
1458 According to Alexander Kolbasov <sasha@unitech.gamma.ru>,
1459 the m4 on UnixWare 2.0 (still in Beta) will core dump on the
1460 config files. GNU m4 and the m4 from UnixWare 1.x both work.
1461
1462 According to Larry Rosenman <ler@lerami.lerctr.org>:
1463
1464 UnixWare 2.1.[23]'s m4 chokes (not obviously) when
1465 processing the 8.9.0 cf files.
1466
1467 I had a LOCAL_RULE_0 that wound up AFTER the
1468 SBasic_check_rcpt rules using the SCO supplied M4.
1469 GNU M4 works fine.
1470
1471UNICOS 8.0.3.4
1472 Some people have reported that the -O flag on UNICOS can cause
1473 problems. You may want to turn this off if you have problems
1474 running sendmail. Reported by Jerry G. DeLapp <jgd@acl.lanl.gov>.
1475
1476Mac OS X (10.0.X)
1477 From: Mike Zimmerman <zimmy@torrentnet.com>
1478 From scratch here is what Darwin users need to do to the standard
1479 10.0.0, 10.0.1 install to get sendmail working.
1480 From http://www.macosx.com/forums/showthread.php?s=6dac0e9e1f3fd118a4870a8a9b559491&threadid=2242:
1481 1. chmod g-w / /private /private/etc
1482 2. Properly set HOSTNAME in /etc/hostconfig to your FQDN:
1483 HOSTNAME=-my.domain.com-
1484 3. Edit /etc/rc.boot:
1485 hostname my.domain.com
1486 domainname domain.com
1487 4. Edit /System/Library/StartupItems/Sendmail/Sendmail:
1488 Remove the "&" after the sendmail command:
1489 /usr/sbin/sendmail -bd -q1h
1490
1491GNU getopt
1492 I'm told that GNU getopt has a problem in that it gets confused
1493 by the double call. Use the version in conf.c instead.
1494
1495BIND 4.9.2 and Ultrix
1496 If you are running on Ultrix, be sure you read conf/Info.Ultrix
1497 in the BIND distribution very carefully -- there is information
1498 in there that you need to know in order to avoid errors of the
1499 form:
1500
1501 /lib/libc.a(gethostent.o): sethostent: multiply defined
1502 /lib/libc.a(gethostent.o): endhostent: multiply defined
1503 /lib/libc.a(gethostent.o): gethostbyname: multiply defined
1504 /lib/libc.a(gethostent.o): gethostbyaddr: multiply defined
1505
1506 during the link stage.
1507
1508BIND 8.X
1509 BIND 8.X returns HOST_NOT_FOUND instead of TRY_AGAIN on temporary
1510 DNS failures when trying to find the hostname associated with an IP
1511 address (gethostbyaddr()). This can cause problems as
1512 $&{client_name} based lookups in class R ($=R) and the access
1513 database won't succeed.
1514
1515 This will be fixed in BIND 8.2.1. For earlier versions, this can
1516 be fixed by making "dns" the last name service queried for host
1517 resolution in /etc/irs.conf:
1518
1519 hosts local continue
1520 hosts dns
1521
1522strtoul
1523 Some compilers (notably gcc) claim to be ANSI C but do not
1524 include the ANSI-required routine "strtoul". If your compiler
1525 has this problem, you will get an error in srvrsmtp.c on the
1526 code:
1527
1528 # ifdef defined(__STDC__) && !defined(BROKEN_ANSI_LIBRARY)
1529 e->e_msgsize = strtoul(vp, (char **) NULL, 10);
1530 # else
1531 e->e_msgsize = strtol(vp, (char **) NULL, 10);
1532 # endif
1533
1534 You can use -DBROKEN_ANSI_LIBRARY to get around this problem.
1535
1536Listproc 6.0c
1537 Date: 23 Sep 1995 23:56:07 GMT
1538 Message-ID: <95925101334.~INN-AUMa00187.comp-news@dl.ac.uk>
1539 From: alansz@mellers1.psych.berkeley.edu (Alan Schwartz)
1540 Subject: Listproc 6.0c + Sendmail 8.7 [Helpful hint]
1541
1542 Just upgraded to sendmail 8.7, and discovered that listproc 6.0c
1543 breaks, because it, by default, sends a blank "HELO" rather than
1544 a "HELO hostname" when using the 'system' or 'telnet' mailmethod.
1545
1546 The fix is to include -DZMAILER in the compilation, which will
1547 cause it to use "HELO hostname" (which Z-mail apparently requires
1548 as well. :)
1549
1550OpenSSL
1551 OpenSSL versions prior to 0.9.6 use a macro named Free which
1552 conflicts with existing macro names on some platforms, such as
1553 AIX.
1554 Do not use 0.9.3, but OpenSSL 0.9.5a or later if compatible with
1555 0.9.5a.
1556
1557sfio
1558 You may run into problems if you use sfio2000 (the body of a
1559 message is lost). Use sfio1999 instead; however, it also has
1560 a bug that can cause sendmail to fail. A patch has been provided
1561 by Petr Lampa of Brno University of Technology, which is given here:
1562
1563diff -rc ../../../../sfio/src/lib/sfio/sfputr.c ./sfputr.c
1564*** ../../../../sfio/src/lib/sfio/sfputr.c Tue May 16 18:25:49 2000
1565--- ./sfputr.c Wed Sep 20 09:06:01 2000
1566***************
1567*** 24,29 ****
1568--- 24,30 ----
1569 for(w = 0; (*s || rc >= 0); )
1570 { SFWPEEK(f,ps,p);
1571
1572+ if(p == -1) return -1; /* PL */
1573 if(p == 0 || (f->flags&SF_WHOLE) )
1574 { n = strlen(s);
1575 if(p >= (n + (rc < 0 ? 0 : 1)) )
1576
1577
1578PH
1579 PH support is provided by Mark Roth <roth@uiuc.edu>. The map is
1580 described at http://www-dev.cso.uiuc.edu/sendmail/ .
1581 Please contact Mark Roth for support and questions regarding the
1582 map.
1583
1584TCP Wrappers
1585 If you are using -DTCPWRAPPERS to get TCP Wrappers support you will
1586 also need to install libwrap.a and modify your site.config.m4 file
1587 or the generated Makefile to include -lwrap in the LIBS line
1588 (make sure that INCDIRS and LIBDIRS point to where the tcpd.h and
1589 libwrap.a can be found).
1590
1591 TCP Wrappers is available at ftp://ftp.porcupine.org/pub/security/.
1592
1593 If you have alternate MX sites for your site, be sure that all of
1594 your MX sites reject the same set of hosts. If not, a bad guy whom
1595 you reject will connect to your site, fail, and move on to the next
1596 MX site, which will accept the mail for you and forward it on to you.
1597
1598Regular Expressions (MAP_REGEX)
1599 If sendmail linking fails with:
1600
1601 undefined reference to 'regcomp'
1602
1603 or sendmail gives an error about a regular expression with:
1604
1605 pattern-compile-error: : Operation not applicable
1606
1607 Your libc does not include a running version of POSIX-regex. Use
1608 librx or regex.o from the GNU Free Software Foundation,
1609 ftp://ftp.gnu.org/pub/gnu/rx-?.?.tar.gz or
1610 ftp://ftp.gnu.org/pub/gnu/regex-?.?.tar.gz.
1611 You can also use the regex-lib by Henry Spencer,
1612 ftp://ftp.funet.fi/pub/languages/C/spencer/regex.shar.gz
1613 Make sure, your compiler reads regex.h from the distribution,
1614 not from /usr/include, otherwise sendmail will dump a core.
1615
1616
1617+--------------+
1618| MANUAL PAGES |
1619+--------------+
1620
1621The manual pages have been written against the -man macros, and
1622should format correctly with any reasonable *roff.
1623
1624
1625+-----------------+
1626| DEBUGGING HOOKS |
1627+-----------------+
1628
1629As of 8.6.5, sendmail daemons will catch a SIGUSR1 signal and log
1630some debugging output (logged at LOG_DEBUG severity). The
1631information dumped is:
1632
1633 * The value of the $j macro.
1634 * A warning if $j is not in the set $=w.
1635 * A list of the open file descriptors.
1636 * The contents of the connection cache.
1637 * If ruleset 89 is defined, it is evaluated and the results printed.
1638
1639This allows you to get information regarding the runtime state of the
1640daemon on the fly. This should not be done too frequently, since
1641the process of rewriting may lose memory which will not be recovered.
1642Also, ruleset 89 may call non-reentrant routines, so there is a small
1643non-zero probability that this will cause other problems. It is
1644really only for debugging serious problems.
1645
1646A typical formulation of ruleset 89 would be:
1647
1648 R$* $@ $>0 some test address
1649
1650
1651+-----------------------------+
1652| DESCRIPTION OF SOURCE FILES |
1653+-----------------------------+
1654
1655The following list describes the files in this directory:
1656
1657Build Shell script for building sendmail.
1658Makefile A convenience for calling ./Build.
1659Makefile.m4 A template for constructing a makefile based on the
1660 information in the devtools directory.
1661README This file.
1662TRACEFLAGS My own personal list of the trace flags -- not guaranteed
1663 to be particularly up to date.
1664alias.c Does name aliasing in all forms.
1665aliases.5 Man page describing the format of the aliases file.
1666arpadate.c A subroutine which creates ARPANET standard dates.
1667bf.h Buffered file I/O function declarations.
1668bf_portable.c Stub routines for systems lacking the Torek stdio library.
1669bf_portable.h Data structure and function declarations for bf_portable.c.
1670bf_torek.c Routines to implement memory-buffered file system using
1671 hooks provided by Torek stdio library.
1672bf_torek.h Data structure and function declarations for bf_torek.c.
1673clock.c Routines to implement real-time oriented functions
1674 in sendmail -- e.g., timeouts.
1675collect.c The routine that actually reads the mail into a temp
1676 file. It also does a certain amount of parsing of
1677 the header, etc.
1678conf.c The configuration file. This contains information
1679 that is presumed to be quite static and non-
1680 controversial, or code compiled in for efficiency
1681 reasons. Most of the configuration is in sendmail.cf.
1682conf.h Configuration that must be known everywhere.
1683convtime.c A routine to sanely process times.
1684daemon.c Routines to implement daemon mode. This version is
1685 specifically for Berkeley 4.1 IPC.
1686deliver.c Routines to deliver mail.
1687domain.c Routines that interface with DNS (the Domain Name
1688 System).
1689envelope.c Routines to manipulate the envelope structure.
1690err.c Routines to print error messages.
1691headers.c Routines to process message headers.
1692helpfile An example helpfile for the SMTP HELP command and -bt mode.
1693macro.c The macro expander. This is used internally to
1694 insert information from the configuration file.
1695mailq.1 Man page for the mailq command.
1696main.c The main routine to sendmail. This file also
1697 contains some miscellaneous routines.
1698makesendmail A convenience for calling ./Build.
1699map.c Support for database maps.
1700mci.c Routines that handle mail connection information caching.
1701milter.c MTA portions of the mail filter API.
1702mime.c MIME conversion routines.
1703newaliases.1 Man page for the newaliases command.
1704parseaddr.c The routines which do address parsing.
1705queue.c Routines to implement message queueing.
1706readcf.c The routine that reads the configuration file and
1707 translates it to internal form.
1708recipient.c Routines that manipulate the recipient list.
1709savemail.c Routines which save the letter on processing errors.
1710sendmail.8 Man page for the sendmail command.
1711sendmail.h Main header file for sendmail.
1712sfsasl.c I/O interface between SASL/TLS and the MTA using SFIO.
1713sfsasl.h Header file for sfsasl.c.
1714shmticklib.c Routines for shared memory counters.
1715srvrsmtp.c Routines to implement server SMTP.
1716stab.c Routines to manage the symbol table.
1717stats.c Routines to collect and post the statistics.
1718statusd_shm.h Data structure and function declarations for shmticklib.c.
1719sysexits.c List of error messages associated with error codes
1720 in sysexits.h.
1721sysexits.h List of error codes for systems that lack their own.
1722timers.c Routines to provide microtimers.
1723timers.h Data structure and function declarations for timers.h.
1724trace.c The trace package. These routines allow setting and
1725 testing of trace flags with a high granularity.
1726udb.c The user database interface module.
1727usersmtp.c Routines to implement user SMTP.
1728util.c Some general purpose routines used by sendmail.
1729version.c The version number and information about this
1730 version of sendmail.
1731
| 118LDAPMAP Lightweight Directory Access Protocol support. You will
119 have to install the UMich or OpenLDAP
120 (http://www.openldap.org/) ldap and lber libraries to use
121 this flag.
122MAP_REGEX Regular Expression support. You will need to use an
123 operating system which comes with the POSIX regex()
124 routines or install a regexp library such as libregex from
125 the Free Software Foundation.
126PH_MAP PH map support. You will need the qi PH package.
127MAP_NSD nsd map support (IRIX 6.5 and later).
128
129>>> NOTE WELL for NEWDB support: If you want to get ndbm support, for
130>>> Berkeley DB versions under 2.0, it is CRITICAL that you remove
131>>> ndbm.o from libdb.a before you install it and DO NOT install ndbm.h;
132>>> for Berkeley DB versions 2.0 through 2.3.14, remove dbm.o from libdb.a
133>>> before you install it. If you don't delete these, there is absolutely
134>>> no point to including -DNDBM, since it will just get you another
135>>> (inferior) API to the same format database. These files OVERRIDE
136>>> calls to ndbm routines -- in particular, if you leave ndbm.h in,
137>>> you can find yourself using the new db package even if you don't
138>>> define NEWDB. Berkeley DB versions later than 2.3.14 do not need
139>>> to be modified. Please also consult the README in the top level
140>>> directory of the sendmail distribution for other important information.
141>>>
142>>> Further note: DO NOT remove your existing /usr/include/ndbm.h --
143>>> you need that one. But do not install an updated ndbm.h in
144>>> /usr/include, /usr/local/include, or anywhere else.
145
146If NEWDB and NDBM are defined (but not NIS), then sendmail will read
147NDBM format alias files, but the next time a newaliases is run the
148format will be converted to NEWDB; that format will be used forever
149more. This is intended as a transition feature.
150
151If NEWDB, NDBM, and NIS are all defined and the name of the file includes
152the string "/yp/", sendmail will rebuild BOTH the NEWDB and NDBM format
153alias files. However, it will only read the NEWDB file; the NDBM format
154file is used only by the NIS subsystem. This is needed because the NIS
155maps on an NIS server are built directly from the NDBM files.
156
157If NDBM and NIS are defined (regardless of the definition of NEWDB),
158and the filename includes the string "/yp/", sendmail adds the special
159tokens "YP_LAST_MODIFIED" and "YP_MASTER_NAME", both of which are
160required if the NDBM file is to be used as an NIS map.
161
162All of these flags are normally defined in the DBMDEF line in the
163Makefile.
164
165If you define NEWDB or HESIOD you get the User Database (USERDB)
166automatically. Generally you do want to have NEWDB for it to do
167anything interesting. See above for getting the Berkeley DB
168package (i.e., NEWDB). There is no separate "user database"
169package -- don't bother searching for it on the net.
170
171Hesiod and LDAP require libraries that may not be installed with your
172system. These are outside of my ability to provide support. See the
173"Quirks" section for more information.
174
175The regex map can be used to see if an address matches a certain regular
176expression. For example, all-numerics local parts are common spam
177addresses, so "^[0-9]+$" would match this. By using such a map in a
178check_* rule-set, you can block a certain range of addresses that would
179otherwise be considered valid.
180
181
182+---------------+
183| COMPILE FLAGS |
184+---------------+
185
186Wherever possible, I try to make sendmail pull in the correct
187compilation options needed to compile on various environments based on
188automatically defined symbols. Some machines don't seem to have useful
189symbols available, requiring that a compilation flag be defined in
190the Makefile; see the devtools/OS subdirectory for the supported
191architectures.
192
193If you are a system to which sendmail has already been ported you
194should not have to touch the following symbols. But if you are porting,
195you may have to tweak the following compilation flags in conf.h in order
196to get it to compile and link properly:
197
198SYSTEM5 Adjust for System V (not necessarily Release 4).
199SYS5SIGNALS Use System V signal semantics -- the signal handler
200 is automatically dropped when the signal is caught.
201 If this is not set, use POSIX/BSD semantics, where the
202 signal handler stays in force until an exec or an
203 explicit delete. Implied by SYSTEM5.
204SYS5SETPGRP Use System V setpgrp() semantics. Implied by SYSTEM5.
205HASFCHMOD Define this to one if you have the fchmod(2) system call.
206 This improves security.
207HASFCHOWN Define this to one if you have the fchown(2) system call.
208 This is required for the TrustedUser option.
209HASFLOCK Set this if you prefer to use the flock(2) system call
210 rather than using fcntl-based locking. Fcntl locking
211 has some semantic gotchas, but many vendor systems
212 also interface it to lockd(8) to do NFS-style locking.
213 Unfortunately, may vendors implementations of fcntl locking
214 is just plain broken (e.g., locks are never released,
215 causing your sendmail to deadlock; when the kernel runs
216 out of locks your system crashes). For this reason, I
217 recommend always defining this unless you are absolutely
218 certain that your fcntl locking implementation really works.
219HASUNAME Set if you have the "uname" system call. Implied by
220 SYSTEM5.
221HASUNSETENV Define this if your system library has the "unsetenv"
222 subroutine.
223HASSETSID Define this if you have the setsid(2) system call. This
224 is implied if your system appears to be POSIX compliant.
225HASINITGROUPS Define this if you have the initgroups(3) routine.
226HASSETVBUF Define this if you have the setvbuf(3) library call.
227 If you don't, setlinebuf will be used instead. This
228 defaults on if your compiler defines __STDC__.
229HASSETREUID Define this if you have setreuid(2) ***AND*** root can
230 use setreuid to change to an arbitrary user. This second
231 condition is not satisfied on AIX 3.x. You may find that
232 your system has setresuid(2), (for example, on HP-UX) in
233 which case you will also have to #define setreuid(r, e)
234 to be the appropriate call. Some systems (such as Solaris)
235 have a compatibility routine that doesn't work properly,
236 but may have "saved user ids" properly implemented so you
237 can ``#define setreuid(r, e) seteuid(e)'' and have it work.
238 The important thing is that you have a call that will set
239 the effective uid independently of the real or saved uid
240 and be able to set the effective uid back again when done.
241 There's a test program in ../test/t_setreuid.c that will
242 try things on your system. Setting this improves the
243 security, since sendmail doesn't have to read .forward
244 and :include: files as root. There are certain attacks
245 that may be unpreventable without this call.
246USESETEUID Define this to 1 if you have a seteuid(2) system call that
247 will allow root to set only the effective user id to an
248 arbitrary value ***AND*** you have saved user ids. This is
249 preferable to HASSETREUID if these conditions are fulfilled.
250 These are the semantics of the to-be-released revision of
251 Posix.1. The test program ../test/t_seteuid.c will try
252 this out on your system. If you define both HASSETREUID
253 and USESETEUID, the former is ignored.
254HASLSTAT Define this if you have symbolic links (and thus the
255 lstat(2) system call). This improves security. Unlike
256 most other options, this one is on by default, so you
257 need to #undef it in conf.h if you don't have symbolic
258 links (these days everyone does).
259HASSETRLIMIT Define this to 1 if you have the setrlimit(2) syscall.
260 You can define it to 0 to force it off. It is assumed
261 if you are running a BSD-like system.
262HASULIMIT Define this if you have the ulimit(2) syscall (System V
263 style systems). HASSETRLIMIT overrides, as it is more
264 general.
265HASWAITPID Define this if you have the waitpid(2) syscall.
266HASGETDTABLESIZE
267 Define this if you have the getdtablesize(2) syscall.
268HAS_ST_GEN Define this to 1 if your system has the st_gen field in
269 the stat structure (see stat(2)).
270HASSRANDOMDEV Define this if your system has the srandomdev(3) function
271 call.
272HASURANDOMDEV Define this if your system has /dev/urandom(4).
273HASSTRERROR Define this if you have the libc strerror(3) function (which
274 should be declared in <errno.h>), and it should be used
275 instead of sys_errlist.
276NEEDGETOPT Define this if you need a reimplementation of getopt(3).
277 On some systems, getopt does very odd things if called
278 to scan the arguments twice. This flag will ask sendmail
279 to compile in a local version of getopt that works
280 properly.
281NEEDSTRTOL Define this if your standard C library does not define
282 strtol(3). This will compile in a local version.
283NEEDVPRINTF Define this if your standard C library does not define
284 vprintf(3). Note that the resulting fake implementation
285 is not very elegant and may not even work on some
286 architectures.
287NEEDFSYNC Define this if your standard C library does not define
288 fsync(2). This will try to simulate the operation using
289 fcntl(2); if that is not available it does nothing, which
290 isn't great, but at least it compiles and runs.
291HASGETUSERSHELL Define this to 1 if you have getusershell(3) in your
292 standard C library. If this is not defined, or is defined
293 to be 0, sendmail will scan the /etc/shells file (no
294 NIS-style support, defaults to /bin/sh and /bin/csh if
295 that file does not exist) to get a list of unrestricted
296 user shells. This is used to determine whether users
297 are allowed to forward their mail to a program or a file.
298NEEDPUTENV Define this if your system needs am emulation of the
299 putenv(3) call. Define to 1 to implement it in terms
300 of setenv(3) or to 2 to do it in terms of primitives.
301NOFTRUNCATE Define this if you don't have the ftruncate(2) syscall.
302 If you don't have this system call, there is an unavoidable
303 race condition that occurs when creating alias databases.
304GIDSET_T The type of entries in a gidset passed as the second
305 argument to getgroups(2). Historically this has been an
306 int, so this is the default, but some systems (such as
307 IRIX) pass it as a gid_t, which is an unsigned short.
308 This will make a difference, so it is important to get
309 this right! However, it is only an issue if you have
310 group sets.
311SLEEP_T The type returned by the system sleep() function.
312 Defaults to "unsigned int". Don't worry about this
313 if you don't have compilation problems.
314ARBPTR_T The type of an arbitrary pointer -- defaults to "void *".
315 If you are an very old compiler you may need to define
316 this to be "char *".
317SOCKADDR_LEN_T The type used for the third parameter to accept(2),
318 getsockname(2), and getpeername(2), representing the
319 length of a struct sockaddr. Defaults to int.
320SOCKOPT_LEN_T The type used for the fifth parameter to getsockopt(2)
321 and setsockopt(2), representing the length of the option
322 buffer. Defaults to int.
323LA_TYPE The type of load average your kernel supports. These
324 can be one of:
325 LA_ZERO (1) -- it always returns the load average as
326 "zero" (and does so on all architectures).
327 LA_INT (2) to read /dev/kmem for the symbol avenrun and
328 interpret as a long integer.
329 LA_FLOAT (3) same, but interpret the result as a floating
330 point number.
331 LA_SHORT (6) to interpret as a short integer.
332 LA_SUBR (4) if you have the getloadavg(3) routine in your
333 system library.
334 LA_MACH (5) to use MACH-style load averages (calls
335 processor_set_info()),
336 LA_PROCSTR (7) to read /proc/loadavg and interpret it
337 as a string representing a floating-point
338 number (Linux-style).
339 LA_READKSYM (8) is an implementation suitable for some
340 versions of SVr4 that uses the MIOC_READKSYM ioctl
341 call to read /dev/kmem.
342 LA_DGUX (9) is a special implementation for DG/UX that uses
343 the dg_sys_info system call.
344 LA_HPUX (10) is an HP-UX specific version that uses the
345 pstat_getdynamic system call.
346 LA_IRIX6 (11) is an IRIX 6.x specific version that adapts
347 to 32 or 64 bit kernels; it is otherwise very similar
348 to LA_INT.
349 LA_KSTAT (12) uses the (Solaris-specific) kstat(3k)
350 implementation.
351 LA_DEVSHORT (13) reads a short from a system file (default:
352 /dev/table/avenrun) and scales it in the same manner
353 as LA_SHORT.
354 LA_INT, LA_SHORT, LA_FLOAT, and LA_READKSYM have several
355 other parameters that they try to divine: the name of your
356 kernel, the name of the variable in the kernel to examine,
357 the number of bits of precision in a fixed point load average,
358 and so forth. LA_DEVSHORT uses _PATH_AVENRUN to find the
359 device to be read to find the load average.
360 In desperation, use LA_ZERO. The actual code is in
361 conf.c -- it can be tweaked if you are brave.
362FSHIFT For LA_INT, LA_SHORT, and LA_READKSYM, this is the number
363 of bits of load average after the binary point -- i.e.,
364 the number of bits to shift right in order to scale the
365 integer to get the true integer load average. Defaults to 8.
366_PATH_UNIX The path to your kernel. Needed only for LA_INT, LA_SHORT,
367 and LA_FLOAT. Defaults to "/unix" on System V, "/vmunix"
368 everywhere else.
369LA_AVENRUN For LA_INT, LA_SHORT, and LA_FLOAT, the name of the kernel
370 variable that holds the load average. Defaults to "avenrun"
371 on System V, "_avenrun" everywhere else.
372SFS_TYPE Encodes how your kernel can locate the amount of free
373 space on a disk partition. This can be set to SFS_NONE
374 (0) if you have no way of getting this information,
375 SFS_USTAT (1) if you have the ustat(2) system call,
376 SFS_4ARGS (2) if you have a four-argument statfs(2)
377 system call (and the include file is <sys/statfs.h>),
378 SFS_VFS (3), SFS_MOUNT (4), SFS_STATFS (5) if you have
379 the two-argument statfs(2) system call with includes in
380 <sys/vfs.h>, <sys/mount.h>, or <sys/statfs.h> respectively,
381 or SFS_STATVFS (6) if you have the two-argument statvfs(2)
382 call. The default if nothing is defined is SFS_NONE.
383SFS_BAVAIL with SFS_4ARGS you can also set SFS_BAVAIL to the field name
384 in the statfs structure that holds the useful information;
385 this defaults to f_bavail.
386SPT_TYPE Encodes how your system can display what a process is doing
387 on a ps(1) command (SPT stands for Set Process Title). Can
388 be set to:
389 SPT_NONE (0) -- Don't try to set the process title at all.
390 SPT_REUSEARGV (1) -- Pad out your argv with the information;
391 this is the default if none specified.
392 SPT_BUILTIN (2) -- The system library has setproctitle.
393 SPT_PSTAT (3) -- Use the PSTAT_SETCMD option to pstat(2)
394 to set the process title; this is used by HP-UX.
395 SPT_PSSTRINGS (4) -- Use the magic PS_STRINGS pointer (4.4BSD).
396 SPT_SYSMIPS (5) -- Use sysmips() supported by NEWS-OS 6.
397 SPT_SCO (6) -- Write kernel u. area.
398 SPT_CHANGEARGV (7) -- Write pointers to our own strings into
399 the existing argv vector.
400SPT_PADCHAR Character used to pad the process title; if undefined,
401 the space character (0x20) is used. This is ignored if
402 SPT_TYPE != SPT_REUSEARGV
403ERRLIST_PREDEFINED
404 If set, assumes that some header file defines sys_errlist.
405 This may be needed if you get type conflicts on this
406 variable -- otherwise don't worry about it.
407WAITUNION The wait(2) routine takes a "union wait" argument instead
408 of an integer argument. This is for compatibility with
409 old versions of BSD.
410SCANF You can set this to extend the F command to accept a
411 scanf string -- this gives you a primitive parser for
412 class definitions -- BUT it can make you vulnerable to
413 core dumps if the target file is poorly formed.
414SYSLOG_BUFSIZE You can define this to be the size of the buffer that
415 syslog accepts. If it is not defined, it assumes a
416 1024-byte buffer. If the buffer is very small (under
417 256 bytes) the log message format changes -- each
418 e-mail message will log many more messages, since it
419 will log each piece of information as a separate line
420 in syslog.
421BROKEN_RES_SEARCH
422 On Ultrix (and maybe other systems?) if you use the
423 res_search routine with an unknown host name, it returns
424 -1 but sets h_errno to 0 instead of HOST_NOT_FOUND. If
425 you set this, sendmail considers 0 to be the same as
426 HOST_NOT_FOUND.
427NAMELISTMASK If defined, values returned by nlist(3) are masked
428 against this value before use -- a common value is
429 0x7fffffff to strip off the top bit.
430BSD4_4_SOCKADDR If defined, socket addresses have an sa_len field that
431 defines the length of this address.
432SAFENFSPATHCONF Set this to 1 if and only if you have verified that a
433 pathconf(2) call with _PC_CHOWN_RESTRICTED argument on an
434 NFS filesystem where the underlying system allows users to
435 give away files to other users returns <= 0. Be sure you
436 try both on NFS V2 and V3. Some systems assume that their
437 local policy apply to NFS servers -- this is a bad
438 assumption! The test/t_pathconf.c program will try this
439 for you -- you have to run it in a directory that is
440 mounted from a server that allows file giveaway.
441SIOCGIFCONF_IS_BROKEN
442 Set this if your system has an SIOCGIFCONF ioctl defined,
443 but it doesn't behave the same way as "most" systems (BSD,
444 Solaris, SunOS, HP-UX, etc.)
445SIOCGIFNUM_IS_BROKEN
446 Set this if your system has an SIOCGIFNUM ioctl defined,
447 but it doesn't behave the same way as "most" systems
448 (Solaris, HP-UX).
449NEED_PERCENTQ Set this if your system doesn't support the printf
450 format strings %lld or %llu. If this is set, %qd and
451 %qu are used instead.
452FAST_PID_RECYCLE
453 Set this if your system can reuse the same PID in the same
454 second.
455SO_REUSEADDR_IS_BROKEN
456 Set this if your system has a setsockopt() SO_REUSEADDR
457 flag but doesn't pay attention to it when trying to bind a
458 socket to a recently closed port.
459SNPRINTF_IS_BROKEN
460 Set this if your system has an snprintf() implementation
461 which does not NUL terminate the string being filled in.
462 Use test/t_snprintf.c to test your system.
463NEEDSGETIPNODE Set this if your system supports IPv6 but doesn't include
464 the getipnodeby{name,addr}() functions. Set automatically
465 for Linux's glibc.
466
467
468+-----------------------+
469| COMPILE-TIME FEATURES |
470+-----------------------+
471
472There are a bunch of features that you can decide to compile in, such
473as selecting various database packages and special protocol support.
474Several are assumed based on other compilation flags -- if you want to
475"un-assume" something, you probably need to edit conf.h. Compilation
476flags that add support for special features include:
477
478NDBM Include support for "new" DBM library for aliases and maps.
479 Normally defined in the Makefile.
480NEWDB Include support for Berkeley DB package (hash & btree)
481 for aliases and maps. Normally defined in the Makefile.
482 If the version of NEWDB you have is the old one that does
483 not include the "fd" call (this call was added in version
484 1.5 of the Berkeley DB code), you must upgrade to the
485 current version of Berkeley DB.
486NIS Define this to get NIS (YP) support for aliases and maps.
487 Normally defined in the Makefile.
488NISPLUS Define this to get NIS+ support for aliases and maps.
489 Normally defined in the Makefile.
490HESIOD Define this to get Hesiod support for aliases and maps.
491 Normally defined in the Makefile.
492NETINFO Define this to get NeXT NetInfo support for aliases and maps.
493 Normally defined in the Makefile.
494LDAPMAP Define this to get LDAP support for maps.
495PH_MAP Define this to get PH support for maps.
496MAP_NSD Define this to get nsd support for maps.
497USERDB Define this to 1 to include support for the User Information
498 Database. Implied by NEWDB or HESIOD. You can use
499 -DUSERDB=0 to explicitly turn it off.
500IDENTPROTO Define this as 1 to get IDENT (RFC 1413) protocol support.
501 This is assumed unless you are running on Ultrix or
502 HP-UX, both of which have a problem in the UDP
503 implementation. You can define it to be 0 to explicitly
504 turn off IDENT protocol support. If defined off, the code
505 is actually still compiled in, but it defaults off; you
506 can turn it on by setting the IDENT timeout in the
507 configuration file.
508IP_SRCROUTE Define this to 1 to get IP source routing information
509 displayed in the Received: header. This is assumed on
510 most systems, but some (e.g., Ultrix) apparently have a
511 broken version of getsockopt that doesn't properly
512 support the IP_OPTIONS call. You probably want this if
513 your OS can cope with it. Symptoms of failure will be that
514 it won't compile properly (that is, no support for fetching
515 IP_OPTIONs), or it compiles but source-routed TCP connections
516 either refuse to open or open and hang for no apparent reason.
517 Ultrix and AIX3 are known to fail this way.
518LOG Set this to get syslog(3) support. Defined by default
519 in conf.h. You want this if at all possible.
520NETINET Set this to get TCP/IP support. Defined by default
521 in conf.h. You probably want this.
522NETINET6 Set this to get IPv6 support. Other configuration may
523 be needed in conf.h for your particular operating system.
524 Also, DaemonPortOptions must be set appropriately for
525 sendmail to accept IPv6 connections.
526NETISO Define this to get ISO networking support.
527NETUNIX Define this to get Unix domain networking support. Defined
528 by default. A few bizarre systems (SCO, ISC, Altos) don't
529 support this networking domain.
530NETNS Define this to get NS networking support.
531NETX25 Define this to get X.25 networking support.
532SMTP Define this to get the SMTP code. Implied by NETINET
533 or NETISO.
534NAMED_BIND If non-zero, include DNS (name daemon) support, including
535 MX support. The specs say you must use this if you run
536 SMTP. You don't have to be running a name server daemon
537 on your machine to need this -- any use of the DNS resolver,
538 including remote access to another machine, requires this
539 option. Defined by default in conf.h. Define it to zero
540 ONLY on machines that do not use DNS in any way.
541QUEUE Define this to get queueing code. Implied by NETINET
542 or NETISO; required by SMTP. This gives you other good
543 stuff -- it should be on.
544DAEMON Define this to get general network support. Implied by
545 NETINET or NETISO. Defined by default in conf.h. You
546 almost certainly want it on.
547MATCHGECOS Permit fuzzy matching of user names against the full
548 name (GECOS) field in the /etc/passwd file. This should
549 probably be on, since you can disable it from the config
550 file if you want to. Defined by default in conf.h.
551MIME8TO7 If non-zero, include 8 to 7 bit MIME conversions. This
552 also controls advertisement of 8BITMIME in the ESMTP
553 startup dialogue.
554MIME7TO8 If non-zero, include 7 to 8 bit MIME conversions.
555HES_GETMAILHOST Define this to 1 if you are using Hesiod with the
556 hes_getmailhost() routine. This is included with the MIT
557 Hesiod distribution, but not with the DEC Hesiod distribution.
558XDEBUG Do additional internal checking. These don't cost too
559 much; you might as well leave this on.
560TCPWRAPPERS Turns on support for the TCP wrappers library (-lwrap).
561 See below for further information.
562SECUREWARE Enable calls to the SecureWare luid enabling/changing routines.
563 SecureWare is a C2 security package added to several UNIX's
564 (notably ConvexOS) to get a C2 Secure system. This
565 option causes mail delivery to be done with the luid of the
566 recipient.
567SHARE_V1 Support for the fair share scheduler, version 1. Setting to
568 1 causes final delivery to be done using the recipients
569 resource limitations. So far as I know, this is only
570 supported on ConvexOS.
571SASL Enables SMTP AUTH (RFC 2554). This requires the Cyrus SASL
572 library (ftp://ftp.andrew.cmu.edu/pub/cyrus-mail/). Please
573 install at least version 1.5.13. See below for further
574 information: SASL COMPILATION AND CONFIGURATION. If your
575 SASL library is older than 1.5.10, you have to set this
576 to its version number using a simple conversion: a.b.c
577 -> c + b*100 + a*10000, e.g. for 1.5.9 define SASL=10509.
578 Note: Using an older version than 1.5.5 of Cyrus SASL is
579 not supported. Starting with version 1.5.10, setting SASL=1
580 is sufficient. Any value other than 1 (or 0) will be
581 compared with the actual version found and if there is a
582 mismatch, compilation will fail.
583EGD Define this if your system has EGD installed, see
584 http://www.lothar.com/tech/crypto/ . It should be used to
585 seed the PRNG for STARTTLS if HASURANDOMDEV is not defined.
586STARTTLS Enables SMTP STARTTLS (RFC 2487). This requires OpenSSL
587 (http://www.OpenSSL.org/) and sfio (see below).
588 Use OpenSSL 0.9.5a or later (if compatible with this
589 version), do not use 0.9.3.
590 See STARTTLS COMPILATION AND CONFIGURATION for further
591 information.
592TLS_NO_RSA Turn off support for RSA algorithms in STARTTLS.
593SFIO Uses sfio instead of stdio. sfio is available from AT&T
594 (http://www.research.att.com/sw/tools/sfio/). If this
595 compile flag is set, confSTDIO_TYPE must be set to portable.
596 This compile flag is necessary for STARTTLS; it also
597 enables the security layer of SASL. The sfio include file
598 stdio.h must be installed in a subdirectory called sfio,
599 i.e., if you install sfio in /usr/local, stdio.h should
600 be in /usr/local/include/sfio, and libsfio.a should be in
601 /usr/local/lib. Notice: read the sfio section in
602 OPERATING SYSTEM AND COMPILE QUIRKS.
603
604
605Generic notice: If you enable a compile time option that needs
606libraries or include files that don't come with sendmail or are
607installed in a location that your C compiler doesn't use by default
608you should set confINCDIRS and confLIBDIRS as explained in the
609first section: BUILDING SENDMAIL.
610
611
612+---------------------+
613| DNS/RESOLVER ISSUES |
614+---------------------+
615
616Many systems have old versions of the resolver library. At a minimum,
617you should be running BIND 4.8.3; older versions may compile, but they
618have known bugs that should give you pause.
619
620Common problems in old versions include "undefined" errors for
621dn_skipname.
622
623Some people have had a problem with BIND 4.9; it uses some routines
624that it expects to be externally defined such as strerror(). It may
625help to link with "-l44bsd" to solve this problem. This has apparently
626been fixed in later versions of BIND, starting around 4.9.3. In other
627words, if you use 4.9.0 through 4.9.2, you need -l44bsd; for earlier or
628later versions, you do not.
629
630!PLEASE! be sure to link with the same version of the resolver as
631the header files you used -- some people have used the 4.9 headers
632and linked with BIND 4.8 or vice versa, and it doesn't work.
633Unfortunately, it doesn't fail in an obvious way -- things just
634subtly don't work.
635
636WILDCARD MX RECORDS ARE A BAD IDEA! The only situation in which they
637work reliably is if you have two versions of DNS, one in the real world
638which has a wildcard pointing to your firewall, and a completely
639different version of the database internally that does not include
640wildcard MX records that match your domain. ANYTHING ELSE WILL GIVE
641YOU HEADACHES!
642
643When attempting to canonify a hostname, some broken name servers will
644return SERVFAIL (a temporary failure) on T_AAAA (IPv6) lookups. If you
645want to excuse this behavior, compile sendmail with
646-D_FFR_WORKAROUND_BROKEN_NAMESERVERS and add WorkAroundBrokenAAAA to your
647ResolverOptions setting. However, instead, we recommend catching the
648problem and reporting it to the name server administrator so we can rid the
649world of broken name servers.
650
651
652+----------------------------------------+
653| STARTTLS COMPILATION AND CONFIGURATION |
654+----------------------------------------+
655
656Please read the docs accompanying the OpenSSL library and sfio.
657You have to compile and install both libraries before you can compile
658sendmail. See devtools/README how to set the correct compile time
659parameters; you should at least set the following variables:
660
661define(`confSTDIO_TYPE', `portable')
662APPENDDEF(`conf_sendmail_ENVDEF', `-DSFIO')
663APPENDDEF(`conf_sendmail_LIBS', `-lsfio')
664APPENDDEF(`conf_sendmail_ENVDEF', `-DSTARTTLS')
665APPENDDEF(`conf_sendmail_LIBS', `-lssl -lcrypto')
666
667If you have installed the OpenSSL libraries and include files in
668a location that your C compiler doesn't use by default you should
669set confINCDIRS and confLIBDIRS as explained in the first section:
670BUILDING SENDMAIL.
671
672Configuration information can be found in doc/op/op.me (required
673certificates) and cf/README (how to tell sendmail about certificates).
674
675To perform an initial test, connect to your sendmail daemon
676(telnet localhost 25) and issue a EHLO localhost and see whether
677250-STARTTLS
678is in the response. If it isn't, run the daemon with
679-O LogLevel=14
680and try again. Then take a look at the logfile and see whether
681there are any problems listed about permissions (unsafe files)
682or the validity of X.509 certificates.
683
684Note: sfio must be used in all libraries with which sendmail exchanges
685file pointers. An example is PH map support. This does not apply to the
686usual libraries, e.g., OpenSSL, Berkeley DB, Cyrus SASL.
687
688Further information can be found via:
689http://www.sendmail.org/tips/
690
691
692+------------------------------------+
693| SASL COMPILATION AND CONFIGURATION |
694+------------------------------------+
695
696Please read the docs accompanying the library (INSTALL and README).
697If you use Berkeley DB for Cyrus SASL then you must compile sendmail
698with the same version of Berkeley DB. See devtools/README how to
699set the correct compile time parameters; you should at least set
700the following variables:
701
702APPENDDEF(`conf_sendmail_ENVDEF', `-DSASL')
703APPENDDEF(`conf_sendmail_LIBS', `-lsasl')
704
705If you have installed the Cyrus SASL library and include files in
706a location that your C compiler doesn't use by default you should
707set confINCDIRS and confLIBDIRS as explained in the first section:
708BUILDING SENDMAIL.
709
710You have to select and install authentication mechanisms and tell
711sendmail where to find the sasl library and the include files (see
712devtools/README for the parameters to set). Setup the required
713users and passwords as explained in the SASL documentation. See
714also cf/README for authentication related options (esp. DefaultAuthInfo
715if you want authentication between MTAs).
716
717To perform an initial test, connect to your sendmail daemon
718(telnet localhost 25) and issue a EHLO localhost and see whether
719250-AUTH ....
720is in the response. If it isn't, run the daemon with
721-O LogLevel=14
722and try again. Then take a look at the logfile and see whether
723there are any security related problems listed (unsafe files).
724
725Further information can be found via:
726http://www.sendmail.org/tips/
727
728
729+-------------------------------------+
730| OPERATING SYSTEM AND COMPILE QUIRKS |
731+-------------------------------------+
732
733GCC problems
734 *****************************************************************
735 ** IMPORTANT: DO NOT USE OPTIMIZATION (``-O'') IF YOU ARE **
736 ** RUNNING GCC 2.4.x or 2.5.x. THERE IS A BUG IN THE GCC **
737 ** OPTIMIZER THAT CAUSES SENDMAIL COMPILES TO FAIL MISERABLY. **
738 *****************************************************************
739
740 Jim Wilson of Cygnus believes he has found the problem -- it will
741 probably be fixed in GCC 2.5.6 -- but until this is verified, be
742 very suspicious of gcc -O. This problem is reported to have been
743 fixed in gcc 2.6.
744
745 A bug in gcc 2.5.5 caused problems compiling sendmail 8.6.5 with
746 optimization on a Sparc. If you are using gcc 2.5.5, youi should
747 upgrade to the latest version of gcc.
748
749 Apparently GCC 2.7.0 on the Pentium processor has optimization
750 problems. I recommend against using -O on that architecture. This
751 has been seen on FreeBSD 2.0.5 RELEASE.
752
753 Solaris 2.X users should use version 2.7.2.3 over 2.7.2.
754
755 We have been told there are problems with gcc 2.8.0. If you are
756 using this version, you should upgrade to 2.8.1 or later.
757
758GDBM GDBM does not work with sendmail 8.8 because the additional
759 security checks and file locking cause problems. Unfortunately,
760 gdbm does not provide a compile flag in its version of ndbm.h so
761 the code can adapt. Until the GDBM authors can fix these problems,
762 GDBM will not be supported. Please use Berkeley DB instead.
763
764Configuration file location
765 Up to 8.6, sendmail tried to find the sendmail.cf file in the same
766 place as the vendors had put it, even when this was obviously
767 stupid. As of 8.7, sendmail ALWAYS looks for /etc/sendmail.cf.
768 Beginning with 8.10, sendmail uses /etc/mail/sendmail.cf.
769 You can get sendmail to use the stupid vendor .cf location by
770 adding -DUSE_VENDOR_CF_PATH during compilation, but this may break
771 support programs and scripts that need to find sendmail.cf. You
772 are STRONGLY urged to use symbolic links if you want to use the
773 vendor location rather than changing the location in the sendmail
774 binary.
775
776 NETINFO systems use NETINFO to determine the location of
777 sendmail.cf. The full path to sendmail.cf is stored as the value of
778 the "sendmail.cf" property in the "/locations/sendmail"
779 subdirectory of NETINFO. Set the value of this property to
780 "/etc/mail/sendmail.cf" (without the quotes) to use this new
781 default location for Sendmail 8.10.0 and higher.
782
783ControlSocket permissions
784 Paraphrased from BIND 8.2.1's README:
785
786 Solaris and other pre-4.4BSD kernels do not respect ownership or
787 protections on UNIX-domain sockets. The short term fix for this is to
788 override the default path and put such control sockets into root-
789 owned directories which do not permit non-root to r/w/x through them.
790 The long term fix is for all kernels to upgrade to 4.4BSD semantics.
791
792SunOS 4.x (Solaris 1.x)
793 You may have to use -lresolv on SunOS. However, beware that
794 this links in a new version of gethostbyname that does not
795 understand NIS, so you must have all of your hosts in DNS.
796
797 Some people have reported problems with the SunOS version of
798 -lresolv and/or in.named, and suggest that you get a newer
799 version. The symptoms are delays when you connect to the
800 SMTP server on a SunOS machine or having your domain added to
801 addresses inappropriately. There is a version of BIND
802 version 4.9 on gatekeeper.DEC.COM in pub/BSD/bind/4.9.
803
804 There is substantial disagreement about whether you can make
805 this work with resolv+, which allows you to specify a search-path
806 of services. Some people report that it works fine, others
807 claim it doesn't work at all (including causing sendmail to
808 drop core when it tries to do multiple resolv+ lookups for a
809 single job). I haven't tried resolv+, as we use DNS exclusively.
810
811 Should you want to try resolv+, it is on ftp.uu.net in
812 /networking/ip/dns.
813
814 Apparently getservbyname() can fail under moderate to high
815 load under some circumstances. This will exhibit itself as
816 the message ``554 makeconnection: service "smtp" unknown''.
817 The problem has been traced to one or more blank lines in
818 /etc/services on the NIS server machine. Delete these
819 and it should work. This info is thanks to Brian Bartholomew
820 <bb@math.ufl.edu> of I-Kinetics, Inc.
821
822 NOTE: The SunOS 4.X linker uses library paths specified during
823 compilation using -L for run-time shared library searches.
824 Therefore, it is vital that relative and unsafe directory paths not
825 be used when compiling sendmail.
826
827SunOS 4.0.2 (Sun 386i)
828 Date: Fri, 25 Aug 1995 11:13:58 +0200 (MET DST)
829 From: teus@oce.nl
830
831 Sendmail 8.7.Beta.12 compiles and runs nearly out of the box with the
832 following changes:
833 * Don't use /usr/5bin in your PATH, but make /usr/5bin/uname
834 available as "uname" command.
835 * Use the defines "-DBSD4_3 -DNAMED_BIND=0" in
836 devtools/OS/SunOS.4.0, which is selected via the "uname" command.
837 I recommend to make available the db-library on the system first
838 (and change the Makefile to use this library).
839 Note that the sendmail.cf and aliases files are found in /etc.
840
841SunOS 4.1.3, 4.1.3_U1
842 Sendmail causes crashes on SunOS 4.1.3 and 4.1.3_U1. According
843 to Sun bug number 1077939:
844
845 If an application does a getsockopt() on a SOCK_STREAM (TCP) socket
846 after the other side of the connection has sent a TCP RESET for
847 the stream, the kernel gets a Bus Trap in the tcp_ctloutput() or
848 ip_ctloutput() routine.
849
850 For 4.1.3, this is fixed in patch 100584-08, available on the
851 Sunsolve 2.7.1 or later CDs. For 4.1.3_U1, this was fixed in patch
852 101790-01 (SunOS 4.1.3_U1: TCP socket and reset problems), later
853 obsoleted by patch 102010-05.
854
855 Sun patch 100584-08 is not currently publicly available on their
856 ftp site but a user has reported it can be found at other sites
857 using a web search engine.
858
859Solaris 2.x (SunOS 5.x)
860 To compile for Solaris, the Makefile built by Build must
861 include a SOLARIS definition which reflects the Solaris version
862 (i.e. -DSOLARIS=20400 for 2.4 or -DSOLARIS=20501 for 2.5.1).
863 If you are using gcc, make sure -I/usr/include is not used (or
864 it might complain about TopFrame). If you are using Sun's cc,
865 make sure /opt/SUNWspro/bin/cc is used instead of /usr/ucb/cc
866 (or it might complain about tm_zone).
867
868 The Solaris "syslog" function is apparently limited to something
869 about 90 characters because of a kernel limitation. If you have
870 source code, you can probably up this number. You can get patches
871 that fix this problem: the patch ids are:
872
873 Solaris 2.1 100834
874 Solaris 2.2 100999
875 Solaris 2.3 101318
876
877 Be sure you have the appropriate patch installed or you won't
878 see system logging.
879
880Solaris 2.4 (SunOS 5.4)
881 If you include /usr/lib at the end of your LD_LIBRARY_PATH you run
882 the risk of getting the wrong libraries under some circumstances.
883 This is because of a new feature in Solaris 2.4, described by
884 Rod.Evans@Eng.Sun.COM:
885
886 >> Prior to SunOS 5.4, any LD_LIBRARY_PATH setting was ignored by the
887 >> runtime linker if the application was setxid (secure), thus your
888 >> applications search path would be:
889 >>
890 >> /usr/local/lib LD_LIBRARY_PATH component - IGNORED
891 >> /usr/lib LD_LIBRARY_PATH component - IGNORED
892 >> /usr/local/lib RPATH - honored
893 >> /usr/lib RPATH - honored
894 >>
895 >> the effect is that path 3 would be the first used, and this would
896 >> satisfy your resolv.so lookup.
897 >>
898 >> In SunOS 5.4 we made the LD_LIBRARY_PATH a little more flexible.
899 >> People who developed setxid applications wanted to be able to alter
900 >> the library search path to some degree to allow for their own
901 >> testing and debugging mechanisms. It was decided that the only
902 >> secure way to do this was to allow a `trusted' path to be used in
903 >> LD_LIBRARY_PATH. The only trusted directory we presently define
904 >> is /usr/lib. Thus a setuid root developer could play with some
905 >> alternative shared object implementations and place them in
906 >> /usr/lib (being root we assume they'ed have access to write in this
907 >> directory). This change was made as part of 1155380 - after a
908 >> *huge* amount of discussion regarding the security aspect of things.
909 >>
910 >> So, in SunOS 5.4 your applications search path would be:
911 >>
912 >> /usr/local/lib from LD_LIBRARY_PATH - IGNORED (untrustworthy)
913 >> /usr/lib from LD_LIBRARY_PATH - honored (trustworthy)
914 >> /usr/local/lib from RPATH - honored
915 >> /usr/lib from RPATH - honored
916 >>
917 >> here, path 2 would be the first used.
918
919Solaris 2.5.1 (SunOS 5.5.1) and 2.6 (SunOS 5.6)
920 Apparently Solaris 2.5.1 patch 103663-01 installs a new
921 /usr/include/resolv.h file that defines the __P macro without
922 checking to see if it is already defined. This new resolv.h is also
923 included in the Solaris 2.6 distribution. This causes compile
924 warnings such as:
925
926 In file included from daemon.c:51:
927 /usr/include/resolv.h:208: warning: `__P' redefined
928 cdefs.h:58: warning: this is the location of the previous definition
929
930 These warnings can be safely ignored or you can create a resolv.h
931 file in the obj.SunOS.5.5.1.* or obj.SunOS.5.6.* directory that reads:
932
933 #undef __P
934 #include "/usr/include/resolv.h"
935
936 Sun is aware of the problem (Sun bug ID 4081053) and it will be fixed
937 in Solaris 2.7.
938
939Solaris 7 (SunOS 5.7)
940 Solaris 7 includes LDAP libraries but the implementation was
941 lacking a few things. The following settings can be placed in
942 devtools/Site/site.SunOS.5.7.m4 if you plan on using those
943 libraries.
944
945 APPENDDEF(`confMAPDEF', `-DLDAPMAP')
946 APPENDDEF(`confENVDEF', `-DLDAP_VERSION_MAX=3')
947 APPENDDEF(`confLIBS', `-lldap')
948
949 Also, Sun's patch 107555 is needed to prevent a crash in the call
950 to ldap_set_option for LDAP_OPT_REFERRALS in ldapmap_setopts if
951 LDAP support is compiled in sendmail.
952
953Solaris
954 If you are using dns for hostname resolution on Solaris, make sure
955 that the 'dns' entry is last on the hosts line in
956 '/etc/nsswitch.conf'. For example, use:
957
958 hosts: nisplus files dns
959
960 Do not use:
961
962 host: nisplus dns [NOTFOUND=return] files
963
964 Note that 'nisplus' above is an illustration. The same comment
965 applies no matter what naming services you are using. If you have
966 anything other than dns last, even after "[NOTFOUND=return]",
967 sendmail may not be able to determine whether an error was
968 temporary or permanent. The error returned by the solaris
969 gethostbyname() is the error for the last lookup used, and other
970 naming services do not have the same concept of temporary failure.
971
972Ultrix
973 By default, the IDENT protocol is turned off on Ultrix. If you
974 are running Ultrix 4.4 or later, or if you have included patch
975 CXO-8919 for Ultrix 4.2 or 4.3 to fix the TCP problem, you can turn
976 IDENT on in the configuration file by setting the "ident" timeout.
977
978 The Ultrix 4.5 Y2K patch (ULTV45-022-1) has changed the resolver
979 included in libc.a. Unfortunately, the __RES symbol hasn't changed
980 and therefore, sendmail can no longer automatically detect the
981 newer version. If you get a compiler error:
982
983 /lib/libc.a(gethostent.o): local_hostname_length: multiply defined
984
985 Then rebuild with this in devtools/Site/site.ULTRIX.m4:
986
987 APPENDDEF(`conf_sendmail_ENVDEF', `-DNEEDLOCAL_HOSTNAME_LENGTH=0')
988
989Digital UNIX (formerly DEC OSF/1)
990 If you are compiling on OSF/1 (DEC Alpha), you must use
991 -L/usr/shlib (otherwise it core dumps on startup). You may also
992 need -mld to get the nlist() function, although some versions
993 apparently don't need this.
994
995 Also, the enclosed makefile removed /usr/sbin/smtpd; if you need
996 it, just create the link to the sendmail binary.
997
998 On DEC OSF/1 3.2 or earlier, the MatchGECOS option doesn't work
999 properly due to a bug in the getpw* routines. If you want to use
1000 this, use -DDEC_OSF_BROKEN_GETPWENT=1. The problem is fixed in 3.2C.
1001
1002 Digital's mail delivery agent, /bin/mail (aka /bin/binmail), will
1003 only preserve the envelope sender in the "From " header if
1004 DefaultUserID is set to daemon. Setting this to mailnull will
1005 cause all mail to have the header "From mailnull ...". To use
1006 a different DefaultUserID, you will need to use a different mail
1007 delivery agent (such as mail.local found in the sendmail
1008 distribution).
1009
1010 On Digital UNIX 4.0 and later, Berkeley DB 1.85 is included with the
1011 operating system and already has the ndbm.o module removed. However,
1012 Digital has modified the original Berkeley DB db.h include file.
1013 This results in the following warning while compiling map.c and udb.c:
1014
1015 cc: Warning: /usr/include/db.h, line 74: The redefinition of the macro
1016 "__signed" conflicts with a current definition because the replacement
1017 lists differ. The redefinition is now in effect.
1018 #define __signed signed
1019 ------------------------^
1020
1021 This warning can be ignored.
1022
1023 Digital UNIX's linker checks /usr/ccs/lib/ before /usr/lib/.
1024 If you have installed a new version of BIND in /usr/include
1025 and /usr/lib, you will experience difficulties as Digital ships
1026 libresolv.a in /usr/ccs/lib/ as well. Be sure to replace both
1027 copies of libresolv.a.
1028
1029IRIX
1030 The header files on SGI IRIX are completely prototyped, and as
1031 a result you can sometimes get some warning messages during
1032 compilation. These can be ignored. There are two errors in
1033 deliver only if you are using gcc, both of the form ``warning:
1034 passing arg N of `execve' from incompatible pointer type''.
1035 Also, if you compile with -DNIS, you will get a complaint
1036 about a declaration of struct dom_binding in a prototype
1037 when compiling map.c; this is not important because the
1038 function being prototyped is not used in that file.
1039
1040 In order to compile sendmail you will have had to install
1041 the developers' option in order to get the necessary include
1042 files.
1043
1044 If you compile with -lmalloc (the fast memory allocator), you may
1045 get warning messages such as the following:
1046
1047 ld32: WARNING 85: definition of _calloc in /usr/lib32/libmalloc.so
1048 preempts that definition in /usr/lib32/mips3/libc.so.
1049 ld32: WARNING 85: definition of _malloc in /usr/lib32/libmalloc.so
1050 preempts that definition in /usr/lib32/mips3/libc.so.
1051 ld32: WARNING 85: definition of _realloc in /usr/lib32/libmalloc.so
1052 preempts that definition in /usr/lib32/mips3/libc.so.
1053 ld32: WARNING 85: definition of _free in /usr/lib32/libmalloc.so
1054 preempts that definition in /usr/lib32/mips3/libc.so.
1055 ld32: WARNING 85: definition of _cfree in /usr/lib32/libmalloc.so
1056 preempts that definition in /usr/lib32/mips3/libc.so.
1057
1058 These are unavoidable and innocuous -- just ignore them.
1059
1060 According to Dave Sill <de5@ornl.gov>, there is a version of the
1061 Berkeley DB library patched to run on Irix 6.2 available from
1062 http://reality.sgi.com/ariel/freeware/#db .
1063
1064IRIX 6.x
1065 If you are using XFS filesystem, avoid using the -32 ABI switch to
1066 the cc compiler if possible.
1067
1068 Broken inet_aton and inet_ntoa on IRIX using gcc: There's
1069 a problem with gcc on IRIX, i.e., gcc can't pass structs
1070 less than 16 bits long unless they are 8 bits; IRIX 6.2 has
1071 some other sized structs. See
1072 http://www.bitmechanic.com/mail-archives/mysql/current/0418.html
1073
1074IRIX 6.4
1075 The IRIX 6.5.4 version of /bin/m4 does not work properly with
1076 sendmail. Either install fw_m4.sw.m4 off the Freeware_May99 CD and
1077 use /usr/freeware/bin/m4 or install and use GNU m4.
1078
1079NeXT or NEXTSTEP
1080 NEXTSTEP 3.3 and earlier ship with the old DBM library. Also,
1081 Berkeley DB does not currently run on NEXTSTEP.
1082
1083 If you are compiling on NEXTSTEP, you will have to create an
1084 empty file "unistd.h" and create a file "dirent.h" containing:
1085
1086 #include <sys/dir.h>
1087 #define dirent direct
1088
1089 (devtools/OS/NeXT should try to do both of these for you.)
1090
1091 Apparently, there is a bug in getservbyname on Nextstep 3.0
1092 that causes it to fail under some circumstances with the
1093 message "SYSERR: service "smtp" unknown" logged. You should
1094 be able to work around this by including the line:
1095
1096 OOPort=25
1097
1098 in your .cf file.
1099
1100BSDI (BSD/386) 1.0, NetBSD 0.9, FreeBSD 1.0
1101 The "m4" from BSDI won't handle the config files properly.
1102 I haven't had a chance to test this myself.
1103
1104 The M4 shipped in FreeBSD and NetBSD 0.9 don't handle the config
1105 files properly. One must use either GNU m4 1.1 or the PD-M4
1106 recently posted in comp.os.386bsd.bugs (and maybe others).
1107 NetBSD-current includes the PD-M4 (as stated in the NetBSD file
1108 CHANGES).
1109
1110 FreeBSD 1.0 RELEASE has uname(2) now. Use -DUSEUNAME in order to
1111 use it (look into devtools/OS/FreeBSD). NetBSD-current may have
1112 it too but it has not been verified.
1113
1114 The latest version of Berkeley DB uses a different naming
1115 scheme than the version that is supplied with your release. This
1116 means you will be able to use the current version of Berkeley DB
1117 with sendmail as long you use the new db.h when compiling
1118 sendmail and link it against the new libdb.a or libdb.so. You
1119 should probably keep the original db.h in /usr/include and the
1120 new db.h in /usr/local/include.
1121
11224.3BSD
1123 If you are running a "virgin" version of 4.3BSD, you'll have
1124 a very old resolver and be missing some header files. The
1125 header files are simple -- create empty versions and everything
1126 will work fine. For the resolver you should really port a new
1127 version (4.8.3 or later) of the resolver; 4.9 is available on
1128 gatekeeper.DEC.COM in pub/BSD/bind/4.9. If you are really
1129 determined to continue to use your old, buggy version (or as
1130 a shortcut to get sendmail working -- I'm sure you have the
1131 best intentions to port a modern version of BIND), you can
1132 copy ../contrib/oldbind.compat.c into sendmail and add the
1133 following to devtools/Site/site.config.m4:
1134
1135 APPENDDEF(`confOBJADD', `oldbind.compat.o')
1136
1137A/UX
1138 Date: Tue, 12 Oct 1993 18:28:28 -0400 (EDT)
1139 From: "Eric C. Hagberg" <hagberg@med.cornell.edu>
1140 Subject: Fix for A/UX ndbm
1141
1142 I guess this isn't really a sendmail bug, however, it is something
1143 that A/UX users should be aware of when compiling sendmail 8.6.
1144
1145 Apparently, the calls that sendmail is using to the ndbm routines
1146 in A/UX 3.0.x contain calls to "broken" routines, in that the
1147 aliases database will break when it gets "just a little big"
1148 (sorry I don't have exact numbers here, but it broke somewhere
1149 around 20-25 aliases for me.), making all aliases non-functional
1150 after exceeding this point.
1151
1152 What I did was to get the gnu-dbm-1.6 package, compile it, and
1153 then re-compile sendmail with "-lgdbm", "-DNDBM", and using the
1154 ndbm.h header file that comes with the gnu-package. This makes
1155 things behave properly.
1156 [NOTE: see comment above about GDBM]
1157
1158 I suppose porting the New Berkeley DB package is another route,
1159 however, I made a quick attempt at it, and found it difficult
1160 (not easy at least); the gnu-dbm package "configured" and
1161 compiled easily.
1162
1163 [NOTE: Berkeley DB version 2.X runs on A/UX and can be used for
1164 database maps.]
1165
1166SCO Unix
1167 From: Thomas Essebier <tom@stallion.oz.au>
1168 Organisation: Stallion Technologies Pty Ltd.
1169
1170 It will probably help those who are trying to configure sendmail 8.6.9
1171 to know that if they are on SCO, they had better set
1172 OI-dnsrch
1173 or they will core dump as soon as they try to use the resolver.
1174 ie. although SCO has _res.dnsrch defined, and is kinda BIND 4.8.3, it
1175 does not inititialise it, nor does it understand 'search' in
1176 /etc/named.boot.
1177 - sigh -
1178
1179 According to SCO, the m4 which ships with UnixWare 2.1.2 is broken.
1180 We recommend installing GNU m4 before attempting to build sendmail.
1181
1182DG/UX
1183 Doug Anderson <dlander@afterlife.ncsc.mil> has successfully run
1184 V8 on the DG/UX 5.4.2 and 5.4R3.x platforms under heavy usage.
1185 Originally, the DG /bin/mail program wasn't compatible with
1186 the V8 sendmail, since the DG /bin/mail requires the environment
1187 variable "_FORCE_MAIL_LOCAL_=yes" be set. Version 8.7 now includes
1188 this in the environment before invoking the local mailer. Some
1189 have used procmail to avoid this problem in the past. It works
1190 but some have experienced file locking problems with their DG/UX
1191 ports of procmail.
1192
1193Apollo DomainOS
1194 If you are compiling on Apollo, you will have to create an empty
1195 file "unistd.h" (for DomainOS 10.3 and earlier) and create a file
1196 "dirent.h" containing:
1197
1198 #include <sys/dir.h>
1199 #define dirent direct
1200
1201 (devtools/OS/DomainOS will attempt to do both of these for you.)
1202
1203HP-UX 8.00
1204 Date: Mon, 24 Jan 1994 13:25:45 +0200
1205 From: Kimmo Suominen <Kimmo.Suominen@lut.fi>
1206 Subject: 8.6.5 w/ HP-UX 8.00 on s300
1207
1208 Just compiled and fought with sendmail 8.6.5 on a HP9000/360 (ie. a
1209 series 300 machine) running HP-UX 8.00.
1210
1211 I was getting segmentation fault when delivering to a local user.
1212 With debugging I saw it was faulting when doing _free@libc... *sigh*
1213 It seems the new implementation of malloc on s300 is buggy as of 8.0,
1214 so I tried out the one in -lmalloc (malloc(3X)). With that it seems
1215 to work just dandy.
1216
1217 When linking, you will get the following error:
1218
1219 ld: multiply defined symbol _freespace in file /usr/lib/libmalloc.a
1220
1221 but you can just ignore it. You might want to add this info to the
1222 README file for the future...
1223
1224Linux
1225 Something broke between versions 0.99.13 and 0.99.14 of Linux:
1226 the flock() system call gives errors. If you are running .14,
1227 you must not use flock. You can do this with -DHASFLOCK=0.
1228
1229 Around the inclusion of bind-4.9.3 & Linux libc-4.6.20, the
1230 initialization of the _res structure changed. If /etc/hosts.conf
1231 was configured as "hosts, bind" the resolver code could return
1232 "Name server failure" errors. This is supposedly fixed in
1233 later versions of libc (>= 4.6.29?), and later versions of
1234 sendmail (> 8.6.10) try to work around the problem.
1235
1236 Some older versions (< 4.6.20?) of the libc/include files conflict
1237 with sendmail's version of cdefs.h. Deleting sendmail's version
1238 on those systems should be non-harmful, and new versions don't care.
1239
1240 Sendmail assumes that libc has snprintf, which has been true since
1241 libc 4.7.0. If you are running an older version, you will need to
1242 use -DHASSNPRINTF=0 in the Makefile. If may be able to use -lbsd
1243 (which includes snprintf) instead of turning this off on versions
1244 of libc between 4.4.4 and 4.7.0 (snprintf improves security, so
1245 you want to use this if at all possible).
1246
1247 NOTE ON LINUX & BIND: By default, the Makefile generated for Linux
1248 includes header files in /usr/local/include and libraries in
1249 /usr/local/lib. If you've installed BIND on your system, the header
1250 files typically end up in the search path and you need to add
1251 "-lresolv" to the LIBS line in your Makefile. Really old versions
1252 may need to include "-l44bsd" as well (particularly if the link phase
1253 complains about missing strcasecmp, strncasecmp or strpbrk).
1254 Complaints about an undefined reference to `__dn_skipname' in
1255 domain.o are a sure sign that you need to add -lresolv to LIBS.
1256 Newer versions of Linux are basically threaded BIND, so you may or
1257 may not see complaints if you accidentally mix BIND
1258 headers/libraries with virginal libc. If you have BIND headers in
1259 /usr/local/include (resolv.h, etc) you *should* be adding -lresolv
1260 to LIBS. Data structures may change and you'd be asking for a
1261 core dump.
1262
1263 A number of problems have been reported regarding the Linux 2.2.0
1264 kernel. So far, these problems have been tracked down to syslog()
1265 and DNS resolution. We believe the problem is with the poll()
1266 implementation in the Linux 2.2.0 kernel and poll()-aware versions
1267 of glib (at least up to 2.0.111).
1268
1269 Some pre-glibc distributions of Linux include a syslog.h that does
1270 not work properly with SFIO. You can fix this by adding
1271 "#include <syslog.h>" to the SFIO version of stdio.h as the very
1272 first line.
1273
1274glibc
1275 glibc 2.2.1 (and possibly other versions) changed the value of
1276 __RES in resolv.h but failed to actually provide the IPv6 API
1277 changes that the change implied. Therefore, compiling with
1278 -DNETINET6 fails.
1279
1280 Workarounds:
1281 1) Compile without -DNETINET6
1282 2) Build against a real BIND 8.2.2 include/lib tree
1283 3) Wait for glibc to fix it
1284
1285AIX 4.X
1286 The AIX 4.X linker uses library paths specified during compilation
1287 using -L for run-time shared library searches. Therefore, it is
1288 vital that relative and unsafe directory paths not be using when
1289 compiling sendmail. Because of this danger, by default, compiles
1290 on AIX use the -blibpath option to limit shared libraries to
1291 /usr/lib and /lib. If you need to allow more directories, such as
1292 /usr/local/lib, modify your devtools/Site/site.AIX.4.2.m4,
1293 site.AIX.4.3.m4, and/or site.AIX.4.x.m4 file(s) and set confLDOPTS
1294 approriately. For example:
1295
1296 define(`confLDOPTS', `-blibpath:/usr/lib:/lib:/usr/local/lib')
1297
1298 Be sure to only add (safe) system directories.
1299
1300 The AIX version of GNU ld also exhibits this problem. If you are
1301 using that version, instead of -blibpath, use its -rpath option.
1302 For example:
1303
1304 gcc -Wl,-rpath /usr/lib -Wl,-rpath /lib -Wl,-rpath /usr/local/lib
1305
1306AIX 4.3.3
1307 From: Valdis.Kletnieks@vt.edu
1308 Date: Sun, 02 Jul 2000 03:58:02 -0400
1309
1310 Under AIX 4.3.3, after applying bos.adt.include 4.3.3.12 to close the
1311 BIND 8.2.2 security holes, you can no longer build with -DNETINET6
1312 because they changed the value of __RES in resolv.h but failed to
1313 actually provide the API changes that the change implied.
1314
1315 Workarounds:
1316 1) Compile without -DNETINET6
1317 2) Build against a real BIND 8.2.2 include/lib tree
1318 3) Wait for IBM to fix it
1319
1320AIX 4.X
1321 The AIX m4 implements a different mechanism for ifdef which is
1322 inconsistent with other versions of m4. Therefore, it will not
1323 work properly with the sendmail Build architecture or m4
1324 configuration method. To work around this problem, please use
1325 GNU m4 from ftp://ftp.gnu.org/pub/gnu/.
1326
1327AIX 3.x
1328 This version of sendmail does not support MB, MG, and MR resource
1329 records, which are supported by AIX sendmail.
1330
1331 Several people have reported that the IBM-supplied named returns
1332 fairly random results -- the named should be replaced. It is not
1333 necessary to replace the resolver, which will simplify installation.
1334 A new BIND resolver can be found at http://www.isc.org/isc/.
1335
1336AIX 3.1.x
1337 The supplied load average code only works correctly for AIX 3.2.x.
1338 For 3.1, use -DLA_TYPE=LA_SUBR and get the latest ``monitor''
1339 package by Jussi Maki <jmaki@hut.fi> from ftp.funet.fi in the
1340 directory pub/unix/AIX/rs6000/monitor-1.12.tar.Z; use the loadavgd
1341 daemon, and the getloadavg subroutine supplied with that package.
1342 If you don't care about load average throttling, just turn off
1343 load average checking using -DLA_TYPE=LA_ZERO.
1344
1345AIX 2.2.1
1346 Date: Mon Dec 4 14:14:56 CST 1995
1347 From: Mark Whetzel <markw@antimatr.houston.tx.us>
1348 Subject: Porting sendmail 8.7.2 to AIX V2 on the RT.
1349
1350 This version of sendmail does not support MB, MG, and MR resource
1351 records, which are supported by AIX sendmail.
1352
1353 AIX V2 on the RT does not have 'paths.h'. Create a null
1354 file in the 'obj' directory to remove this compile error.
1355
1356 A patch file is needed to get the BSD 'db' library to compile
1357 for AIX/RT. I have sent the necessary updates to the author,
1358 but they may not be immediately available.
1359 [NOTE: Berkeley DB version 2.X runs on AIX/RT.]
1360
1361 The original AIX/RT resolver libraries are very old, and you
1362 should get the latest BIND to replace it. The 4.8.3 version
1363 has been tested, but 4.9.x is out and should work.
1364
1365 To make the load average code work correctly requires an
1366 external routine, as the kernel does not maintain system
1367 load averages, similar to AIX V3.1.x. A reverse port of the
1368 older 1.05 'monitor' load average daemon code written by
1369 Jussi Maki that will work on AIX V2 for the RT is available
1370 by E-mail to Mark Whetzel <markw@antimatr.houston.tx.us>.
1371 That code depends on an external daemon to collect system
1372 load information, and the external routine 'getloadavg',
1373 that will return that information. The 'LA_SUBR' define
1374 will handle this for AIX V2 on the RT.
1375
1376 Note: You will have to change devtools/OS/AIX.2 to correctly
1377 point to the locatons of the updated BIND source tree and
1378 the location of the 'newdb' tree and library location.
1379 You will also have to change devtools/OS/AIX.2 to know
1380 about the location of the 'getloadavg' routine if you use
1381 the LA_SUBR define.
1382
1383RISC/os
1384 RISC/os from MIPS is a merged AT&T/Berkeley system. When you
1385 compile on that platform you will get duplicate definitions
1386 on many files. You can ignore these.
1387
1388System V Release 4 Based Systems
1389 There is a single devtools OS that is intended for all SVR4-based
1390 systems (built from devtools/OS/SVR4). It defines __svr4__,
1391 which is predefined by some compilers. If your compiler already
1392 defines this compile variable, you can delete the definition from
1393 the generated Makefile or create a devtools/Site/site.config.m4
1394 file.
1395
1396 It's been tested on Dell Issue 2.2.
1397
1398DELL SVR4
1399 Date: Mon, 06 Dec 1993 10:42:29 EST
1400 From: "Kimmo Suominen" <kim@grendel.lut.fi>
1401 Message-ID: <2d0352f9.lento29@lento29.UUCP>
1402 To: eric@cs.berkeley.edu
1403 Cc: sendmail@cs.berkeley.edu
1404 Subject: Notes for DELL SVR4
1405
1406 Eric,
1407
1408 Here are some notes for compiling Sendmail 8.6.4 on DELL SVR4. I ran
1409 across these things when helping out some people who contacted me by
1410 e-mail.
1411
1412 1) Use gcc 2.4.5 (or later?). Dell distributes gcc 2.1 with their
1413 Issue 2.2 Unix. It is too old, and gives you problems with
1414 clock.c, because sigset_t won't get defined in <sys/signal.h>.
1415 This is due to a problematic protection rule in there, and is
1416 fixed with gcc 2.4.5.
1417
1418 2) If you don't use the new Berkeley DB (-DNEWDB), then you need
1419 to add "-lc -lucb" to the libraries to link with. This is because
1420 the -ldbm distributed by Dell needs the bcopy, bcmp and bzero
1421 functions. It is important that you specify both libraries in
1422 the given order to be sure you only get the BSTRING functions
1423 from the UCB library (and not the signal routines etc.).
1424
1425 3) Don't leave out "-lelf" even if compiling with "-lc -lucb".
1426 The UCB library also has another copy of the nlist routines,
1427 but we do want the ones from "-lelf".
1428
1429 If anyone needs a compiled gcc 2.4.5 and/or a ported DB library, they
1430 can use anonymous ftp to fetch them from lut.fi in the /kim directory.
1431 They are copies of what I use on grendel.lut.fi, and offering them
1432 does not imply that I would also support them. I have sent the DB
1433 port for SVR4 back to Keith Bostic for inclusion in the official
1434 distribution, but I haven't heard anything from him as of today.
1435
1436 - gcc-2.4.5-svr4.tar.gz (gcc 2.4.5 and the corresponding libg++)
1437 - db-1.72.tar.gz (with source, objects and a installed copy)
1438
1439 Cheers
1440 + Kim
1441 --
1442 * Kimmo.Suominen@lut.fi * SysVr4 enthusiast at GRENDEL.LUT.FI *
1443 * KIM@FINFILES.BITNET * Postmaster and Hostmaster at LUT.FI *
1444 * + 358 200 865 718 * Unix area moderator at NIC.FUNET.FI *
1445
1446ConvexOS 10.1 and below
1447 In order to use the name server, you must create the file
1448 /etc/use_nameserver. If this file does not exist, the call
1449 to res_init() will fail and you will have absolutely no
1450 access to DNS, including MX records.
1451
1452Amdahl UTS 2.1.5
1453 In order to get UTS to work, you will have to port BIND 4.9.
1454 The vendor's BIND is reported to be ``totally inadequate.''
1455 See sendmail/contrib/AmdahlUTS.patch for the patches necessary
1456 to get BIND 4.9 compiled for UTS.
1457
1458UnixWare
1459 According to Alexander Kolbasov <sasha@unitech.gamma.ru>,
1460 the m4 on UnixWare 2.0 (still in Beta) will core dump on the
1461 config files. GNU m4 and the m4 from UnixWare 1.x both work.
1462
1463 According to Larry Rosenman <ler@lerami.lerctr.org>:
1464
1465 UnixWare 2.1.[23]'s m4 chokes (not obviously) when
1466 processing the 8.9.0 cf files.
1467
1468 I had a LOCAL_RULE_0 that wound up AFTER the
1469 SBasic_check_rcpt rules using the SCO supplied M4.
1470 GNU M4 works fine.
1471
1472UNICOS 8.0.3.4
1473 Some people have reported that the -O flag on UNICOS can cause
1474 problems. You may want to turn this off if you have problems
1475 running sendmail. Reported by Jerry G. DeLapp <jgd@acl.lanl.gov>.
1476
1477Mac OS X (10.0.X)
1478 From: Mike Zimmerman <zimmy@torrentnet.com>
1479 From scratch here is what Darwin users need to do to the standard
1480 10.0.0, 10.0.1 install to get sendmail working.
1481 From http://www.macosx.com/forums/showthread.php?s=6dac0e9e1f3fd118a4870a8a9b559491&threadid=2242:
1482 1. chmod g-w / /private /private/etc
1483 2. Properly set HOSTNAME in /etc/hostconfig to your FQDN:
1484 HOSTNAME=-my.domain.com-
1485 3. Edit /etc/rc.boot:
1486 hostname my.domain.com
1487 domainname domain.com
1488 4. Edit /System/Library/StartupItems/Sendmail/Sendmail:
1489 Remove the "&" after the sendmail command:
1490 /usr/sbin/sendmail -bd -q1h
1491
1492GNU getopt
1493 I'm told that GNU getopt has a problem in that it gets confused
1494 by the double call. Use the version in conf.c instead.
1495
1496BIND 4.9.2 and Ultrix
1497 If you are running on Ultrix, be sure you read conf/Info.Ultrix
1498 in the BIND distribution very carefully -- there is information
1499 in there that you need to know in order to avoid errors of the
1500 form:
1501
1502 /lib/libc.a(gethostent.o): sethostent: multiply defined
1503 /lib/libc.a(gethostent.o): endhostent: multiply defined
1504 /lib/libc.a(gethostent.o): gethostbyname: multiply defined
1505 /lib/libc.a(gethostent.o): gethostbyaddr: multiply defined
1506
1507 during the link stage.
1508
1509BIND 8.X
1510 BIND 8.X returns HOST_NOT_FOUND instead of TRY_AGAIN on temporary
1511 DNS failures when trying to find the hostname associated with an IP
1512 address (gethostbyaddr()). This can cause problems as
1513 $&{client_name} based lookups in class R ($=R) and the access
1514 database won't succeed.
1515
1516 This will be fixed in BIND 8.2.1. For earlier versions, this can
1517 be fixed by making "dns" the last name service queried for host
1518 resolution in /etc/irs.conf:
1519
1520 hosts local continue
1521 hosts dns
1522
1523strtoul
1524 Some compilers (notably gcc) claim to be ANSI C but do not
1525 include the ANSI-required routine "strtoul". If your compiler
1526 has this problem, you will get an error in srvrsmtp.c on the
1527 code:
1528
1529 # ifdef defined(__STDC__) && !defined(BROKEN_ANSI_LIBRARY)
1530 e->e_msgsize = strtoul(vp, (char **) NULL, 10);
1531 # else
1532 e->e_msgsize = strtol(vp, (char **) NULL, 10);
1533 # endif
1534
1535 You can use -DBROKEN_ANSI_LIBRARY to get around this problem.
1536
1537Listproc 6.0c
1538 Date: 23 Sep 1995 23:56:07 GMT
1539 Message-ID: <95925101334.~INN-AUMa00187.comp-news@dl.ac.uk>
1540 From: alansz@mellers1.psych.berkeley.edu (Alan Schwartz)
1541 Subject: Listproc 6.0c + Sendmail 8.7 [Helpful hint]
1542
1543 Just upgraded to sendmail 8.7, and discovered that listproc 6.0c
1544 breaks, because it, by default, sends a blank "HELO" rather than
1545 a "HELO hostname" when using the 'system' or 'telnet' mailmethod.
1546
1547 The fix is to include -DZMAILER in the compilation, which will
1548 cause it to use "HELO hostname" (which Z-mail apparently requires
1549 as well. :)
1550
1551OpenSSL
1552 OpenSSL versions prior to 0.9.6 use a macro named Free which
1553 conflicts with existing macro names on some platforms, such as
1554 AIX.
1555 Do not use 0.9.3, but OpenSSL 0.9.5a or later if compatible with
1556 0.9.5a.
1557
1558sfio
1559 You may run into problems if you use sfio2000 (the body of a
1560 message is lost). Use sfio1999 instead; however, it also has
1561 a bug that can cause sendmail to fail. A patch has been provided
1562 by Petr Lampa of Brno University of Technology, which is given here:
1563
1564diff -rc ../../../../sfio/src/lib/sfio/sfputr.c ./sfputr.c
1565*** ../../../../sfio/src/lib/sfio/sfputr.c Tue May 16 18:25:49 2000
1566--- ./sfputr.c Wed Sep 20 09:06:01 2000
1567***************
1568*** 24,29 ****
1569--- 24,30 ----
1570 for(w = 0; (*s || rc >= 0); )
1571 { SFWPEEK(f,ps,p);
1572
1573+ if(p == -1) return -1; /* PL */
1574 if(p == 0 || (f->flags&SF_WHOLE) )
1575 { n = strlen(s);
1576 if(p >= (n + (rc < 0 ? 0 : 1)) )
1577
1578
1579PH
1580 PH support is provided by Mark Roth <roth@uiuc.edu>. The map is
1581 described at http://www-dev.cso.uiuc.edu/sendmail/ .
1582 Please contact Mark Roth for support and questions regarding the
1583 map.
1584
1585TCP Wrappers
1586 If you are using -DTCPWRAPPERS to get TCP Wrappers support you will
1587 also need to install libwrap.a and modify your site.config.m4 file
1588 or the generated Makefile to include -lwrap in the LIBS line
1589 (make sure that INCDIRS and LIBDIRS point to where the tcpd.h and
1590 libwrap.a can be found).
1591
1592 TCP Wrappers is available at ftp://ftp.porcupine.org/pub/security/.
1593
1594 If you have alternate MX sites for your site, be sure that all of
1595 your MX sites reject the same set of hosts. If not, a bad guy whom
1596 you reject will connect to your site, fail, and move on to the next
1597 MX site, which will accept the mail for you and forward it on to you.
1598
1599Regular Expressions (MAP_REGEX)
1600 If sendmail linking fails with:
1601
1602 undefined reference to 'regcomp'
1603
1604 or sendmail gives an error about a regular expression with:
1605
1606 pattern-compile-error: : Operation not applicable
1607
1608 Your libc does not include a running version of POSIX-regex. Use
1609 librx or regex.o from the GNU Free Software Foundation,
1610 ftp://ftp.gnu.org/pub/gnu/rx-?.?.tar.gz or
1611 ftp://ftp.gnu.org/pub/gnu/regex-?.?.tar.gz.
1612 You can also use the regex-lib by Henry Spencer,
1613 ftp://ftp.funet.fi/pub/languages/C/spencer/regex.shar.gz
1614 Make sure, your compiler reads regex.h from the distribution,
1615 not from /usr/include, otherwise sendmail will dump a core.
1616
1617
1618+--------------+
1619| MANUAL PAGES |
1620+--------------+
1621
1622The manual pages have been written against the -man macros, and
1623should format correctly with any reasonable *roff.
1624
1625
1626+-----------------+
1627| DEBUGGING HOOKS |
1628+-----------------+
1629
1630As of 8.6.5, sendmail daemons will catch a SIGUSR1 signal and log
1631some debugging output (logged at LOG_DEBUG severity). The
1632information dumped is:
1633
1634 * The value of the $j macro.
1635 * A warning if $j is not in the set $=w.
1636 * A list of the open file descriptors.
1637 * The contents of the connection cache.
1638 * If ruleset 89 is defined, it is evaluated and the results printed.
1639
1640This allows you to get information regarding the runtime state of the
1641daemon on the fly. This should not be done too frequently, since
1642the process of rewriting may lose memory which will not be recovered.
1643Also, ruleset 89 may call non-reentrant routines, so there is a small
1644non-zero probability that this will cause other problems. It is
1645really only for debugging serious problems.
1646
1647A typical formulation of ruleset 89 would be:
1648
1649 R$* $@ $>0 some test address
1650
1651
1652+-----------------------------+
1653| DESCRIPTION OF SOURCE FILES |
1654+-----------------------------+
1655
1656The following list describes the files in this directory:
1657
1658Build Shell script for building sendmail.
1659Makefile A convenience for calling ./Build.
1660Makefile.m4 A template for constructing a makefile based on the
1661 information in the devtools directory.
1662README This file.
1663TRACEFLAGS My own personal list of the trace flags -- not guaranteed
1664 to be particularly up to date.
1665alias.c Does name aliasing in all forms.
1666aliases.5 Man page describing the format of the aliases file.
1667arpadate.c A subroutine which creates ARPANET standard dates.
1668bf.h Buffered file I/O function declarations.
1669bf_portable.c Stub routines for systems lacking the Torek stdio library.
1670bf_portable.h Data structure and function declarations for bf_portable.c.
1671bf_torek.c Routines to implement memory-buffered file system using
1672 hooks provided by Torek stdio library.
1673bf_torek.h Data structure and function declarations for bf_torek.c.
1674clock.c Routines to implement real-time oriented functions
1675 in sendmail -- e.g., timeouts.
1676collect.c The routine that actually reads the mail into a temp
1677 file. It also does a certain amount of parsing of
1678 the header, etc.
1679conf.c The configuration file. This contains information
1680 that is presumed to be quite static and non-
1681 controversial, or code compiled in for efficiency
1682 reasons. Most of the configuration is in sendmail.cf.
1683conf.h Configuration that must be known everywhere.
1684convtime.c A routine to sanely process times.
1685daemon.c Routines to implement daemon mode. This version is
1686 specifically for Berkeley 4.1 IPC.
1687deliver.c Routines to deliver mail.
1688domain.c Routines that interface with DNS (the Domain Name
1689 System).
1690envelope.c Routines to manipulate the envelope structure.
1691err.c Routines to print error messages.
1692headers.c Routines to process message headers.
1693helpfile An example helpfile for the SMTP HELP command and -bt mode.
1694macro.c The macro expander. This is used internally to
1695 insert information from the configuration file.
1696mailq.1 Man page for the mailq command.
1697main.c The main routine to sendmail. This file also
1698 contains some miscellaneous routines.
1699makesendmail A convenience for calling ./Build.
1700map.c Support for database maps.
1701mci.c Routines that handle mail connection information caching.
1702milter.c MTA portions of the mail filter API.
1703mime.c MIME conversion routines.
1704newaliases.1 Man page for the newaliases command.
1705parseaddr.c The routines which do address parsing.
1706queue.c Routines to implement message queueing.
1707readcf.c The routine that reads the configuration file and
1708 translates it to internal form.
1709recipient.c Routines that manipulate the recipient list.
1710savemail.c Routines which save the letter on processing errors.
1711sendmail.8 Man page for the sendmail command.
1712sendmail.h Main header file for sendmail.
1713sfsasl.c I/O interface between SASL/TLS and the MTA using SFIO.
1714sfsasl.h Header file for sfsasl.c.
1715shmticklib.c Routines for shared memory counters.
1716srvrsmtp.c Routines to implement server SMTP.
1717stab.c Routines to manage the symbol table.
1718stats.c Routines to collect and post the statistics.
1719statusd_shm.h Data structure and function declarations for shmticklib.c.
1720sysexits.c List of error messages associated with error codes
1721 in sysexits.h.
1722sysexits.h List of error codes for systems that lack their own.
1723timers.c Routines to provide microtimers.
1724timers.h Data structure and function declarations for timers.h.
1725trace.c The trace package. These routines allow setting and
1726 testing of trace flags with a high granularity.
1727udb.c The user database interface module.
1728usersmtp.c Routines to implement user SMTP.
1729util.c Some general purpose routines used by sendmail.
1730version.c The version number and information about this
1731 version of sendmail.
1732
|