cksum -c: support checking base64 digests in reverse mode too.

In reverse mode, the entry does not include the digest algorithm
so the specified algorithm (or the default) is used when checking.
OK deraadt@

Add explicit casts to double to quiet a clang warning.
OK deraddt@

Use ferror() to check for write error, not the fflush() return value.
We can't rely on buffering to catch write errors in fflush().
Based on a diff from Demi M. Obenour, OK kn@

freezero() is of no use here, the computed digest is not a secret.
ok millert@

I am retiring my old email address; replace it with my OpenBSD one.

fgetln(3) -> getline(3); from Lauri Tirkkonen; ok millert@

Use getrusage(2) for the built-in time trial instead of getttimeofday(2).
This makes it possible to measure just the ellapsed user time, which
is what we really care about when benchmarking an algorithm.
OK deraadt@

obvious use for freezero()

In -C mode, process all checksums that match the specified file(s),
not just the first one that matches. OK deraadt@

Move declaration of an 'i' inside the scope of its use. Which in
turn is inside an #ifdef. Thus making the code clearer by eliminating
an #ifdef in the middle of the declarations.

Suggested by millert@

Eliminate some gcc warnings about 'unused variables', mostly by
adding appropriate #ifdef's around declarations.

ok millert@ (with a tweak I will commit separately)

add sha512/256 support to check that the libc code is right :)

Pull in <sys/time.h> for gettimeofday()

all of the update functions take a size_t. correct type and casts.
ok deraadt millert

Remove NULL-check before free(). ok tb@

normalize the ordering of tame requests (particularily, "rpath wpath cpath",
which i have put in that order). this is not important, but helps look
for outliers which might be strange. it hints that "ioctl" should be
reassessed in a few places, to see if "tty" is better; that "unix" may
be used in some places where "route" could now work.

Change all tame callers to namechange to pledge(2).

Repair tame() error check to be == -1

right at startup, this can tame "stdio cpath rpath wpath". after getopt
-h has handled write/creating a file, we can drop to tame "stdio rpath"
since md5 will only read files after that.
i believe i involved lteo for this.

cleanup of the SMALL hiding from Mathias Svensson

Replace <sys/param.h> with <limits.h> and other less dirty headers where
possible. Annotate <sys/param.h> lines with their current reasons. Switch
to PATH_MAX, NGROUPS_MAX, HOST_NAME_MAX+1, LOGIN_NAME_MAX, etc. Change
MIN() and MAX() to local definitions of MINIMUM() and MAXIMUM() where
sensible to avoid pulling in the pollution. These are the files confirmed
through binary verification.
ok guenther, millert, doug (helped with the verification protocol)

Replace all queue *_END macro calls except CIRCLEQ_END with NULL.

CIRCLEQ_* is deprecated and not called in the tree. The other queue types
have *_END macros which were added for symmetry with CIRCLEQ_END. They are
defined as NULL. There's no reason to keep the other *_END macro calls.

ok millert@

Fix memory leak in digest_file() on ferror(). OK tedu@ lteo@

Ignore the -b option if cksum is called as-is (e.g. "cksum -b /bsd")
as documented on the cksum(1) man page.

"looks right" deraadt@
ok millert@

purge sum and sysvsum code. ok deraadt jmc lteo millert
a few initial leftovers spotted by naddy

remove insecure MD4 checksum algorithm; ok tedu@, man page ok jmc@

"A collision attack published in 2007 can find collisions for full
MD4 in less than two hash operations."

When using the -C option, exit with an exit status of 1 if any of the
files specified on the command line do not exist in the checklist.

ok deraadt@ tedu@

add -C to the man page, and adjust usage();
ok lteo

Add support for a -C option. It works on a checklist like -c but only
does the checksum comparison for selected files that are specified on
the command line.

idea discussed with deraadt@ and tedu@
manpage feedback jmc@
feedback/OK deraadt@ millert@

Also move case 'c' into the #ifdef for a smaller binary. It will fall into
default, giving a nice failure. I have not removed -c from the usage()
or getopt() because it is too much butchering...

use -DSHA2_ONLY to be more clear about what we are butchering
idea from tedu

catch up to the fact that md5/sha* got merged, and document -c consistently;
some style and cleanup tweaks while here

ok deraadt

When using a checklist, print MISSING for non-existent files.

Based on an earlier diff by tedu@
Requested by deraadt@
OK deraadt@

the -c option is really a mode change, incompatible with other options.
(there are some others too, but -c is particularly misleading.) split it
out in synopsis and usage.
ok deraadt

Fix wrapping of usage message for cksum.

err() not errx() when fopen fails.

Remove NHASHES, it is not needed and is incompatible with #ifdef SMALL.
Remove style_hash and style_cksum, adjust the style of the selected
algorithms based on rflag and qflag instead.

Remove pmode, fix hfile open mode, check return value when writing
file name to stdout.

Add -DSMALL support to kill a lot of the unncecessary hashes and checksums
that the install media won't need
ok millert, but he used the word butcher!

Add support for -h hashfile. This is very useful with with -p, in a pipeline,
to deliver the hash output to a different location.
ok millert

install a link as sha512. simplify program internals; there are only
two modes. ok deraadt gilles jca

add unsigned char casts for specific calls to ctype.h macros.
ok guenther step

SHA-224 is to SHA-256 as SHA-384 is to SHA-512. It was added in a
later revision of FIPS-180. OK miod@ jmc@ guenther@ djm@

Return an exit code of 1 if the file cannot be opened (e.g. file does
not exist), or if there is an error reading the file stream.

From Patrik Lundin, thanks!

ok deraadt

replace sys/param.h with sys/types.h (and sometimes add limits.h if needed)
ok guenther

fix memset sizeof, found by jsg. ok krw

Use a 32KB data buffer instead of a 1KB buffer.

Use stdio when reading in the file to be checksummed. This can improve
performance due to the internal buffering stdio does wrt the optimal
read size (st_blksize). OK mikeb@

adapt to sha2(3) API changes; ok millert@

md5(1), sha1(1), rmd160(1), cksum(1) and sum(1) have different options;
while here, some KNF improvements.

based on a diff for compress(1) written by millert@,
fixes from guenther@ and pyr@.

ok millert@, pyr@; (for the manual page tweaks) jmc@

properly exit >0 when encountering bad md5 in cheklist

ok gilles@ millert@

From Tim van der Molen <tbvdm.lists at xs4all.nl>

clamp -ttttttttt test counter to 1B, so that it does not wrap negative;
from Nicholas Marriott, ok otto

if we define a constant for the OPTSTRING, we can as well use it.

ok tedu@

allow multiple -t to increase the test count. cpus are getting too fast.
ok grunk. with manpage from grunk and jmc

Fix a cust & pasto
Correctly compute amount of base64 padding in checklist mode

Do not print a warning that no valid checksums were found if there
was a single fingerprint for a file that could not be opened.
The warning that the file could not be opened is sufficient.
From espie@

Supress warnings for the first getopt() pass. Otherwise,
any "unknown option" messages get printed twice.

Add base64 support to cksum and friends. Output encoding can
be set globally or on a per-algorithm basis when using multiple
hash functions. OK espie@ simon@ deraadt@; man help from jmc@

augment the -q option, now if specifed with -c (checklist)
it will only print out the failed cases.

OK millert@

align usage message correctly

simplify synopsis and usage(); ok millert

Add -r and -q options similar to what is supported by FreeBSD's md5
OK deraadt@ ckuethe@

Accept upper and lower case letters in a checksum, since this doesn't
alter the value of the digest. Nor, for that matter, does case affect
the name of the algorithm used.

Partly from Matthew Clarke, clamat (at) telus (dot) net.

ok djm@ millert@

Fix signed/unsigned comparisons by using size_t
Sprinkle some const

ok ray@ jaredy@

Make "cksum -c" grok checklists generated with cksum, sum and sysvsum.
OK deraadt@

Exit unsuccessfully when comparing against a checklist file fails.
OK markus@ millert@

Add support for multiple using algorithms. OK'd by several people
a while ago so get this out of my tree...

GNU md5sum uses tab not space to separate filename and hash. Fixes
parsing of GNU-style checklist files for files with spaces in them.
From Emil Mikulic; closes PR 4027

sync usage w/ man page

Add support for cksum (3 varieties), md4, sha256, sha384, and sha512.
Consequently, cksum moves becomes a link to md5 and moves to /bin.
The cksum man page could use some polishing.

Use the FOO_DIGEST_LENGTH macros

Extend "md5 -c" to parse the output of GNU md5sum.
Adapted from patches by Lawrence Teo and tedu@. OK tedu@

Fix typo: unsigned -> unsigned char; from Thorsten Glaser

fix lots of proto issues

Sync with share/misc/license.template and add missing DARPA credit
where applicable.

ansification, art ok

Use an ISC-tyle license for all my code; it is simpler and more permissive.

return from main

strchr() should be strrchr(); zyrnix

Do not closed stdin; Andrey Matveev

Cast 3 to size_t to shut up stupid gcc on 64bit platforms.

Add -c option similar to GNU md5sum.
Original patch from zyrnix but significantly modified by me.
deraadt@ OK

Report read() errors; noticed and OK by fgs@

Swap TEST_BLOCK_LEN and TEST_BLOCK_COUNT in benchmark printf.

ok millert@

o) __progname aria;

millert@ ok.

Nope, that's not right either. Fix check for mutually exclusive args.

Obvious thinko: command line is wrong if it's too short, not too long.

Fix some typos and add missing test for string of 1 million a's.

New md5/sha1/rmd160 driver with a BSD copyright. Improvements:
o options that should be mutually exclusive are
o time trial works reasonably when it finishes in < 1 sec
o uses a function table instead of a bunch of globals

One single malloc, with a union to get the largest needed size,
should be enough.

use getopt; mph@pobox.com

warn not warnx

Adds rmd160 functionality.

md5(1) is now capable of doing md4, md5 and sha1 digests. Currently
only md5 and sha1 are used.

boost TEST_BLOCK_LEN and TEST_BLOCK_COUNT a bit. this makes '-t' more
useful on fast boxes.
also fflush(stdout) after printing the first test message so
the user knows its actually doing something

64-bit cleanup + pedantic -W flags

-Wall happiness

move to bin/md5 for `greater accessibility'

Add explicit casts to double to quiet a clang warning.
OK deraddt@

Use ferror() to check for write error, not the fflush() return value.
We can't rely on buffering to catch write errors in fflush().
Based on a diff from Demi M. Obenour, OK kn@

freezero() is of no use here, the computed digest is not a secret.
ok millert@

I am retiring my old email address; replace it with my OpenBSD one.

fgetln(3) -> getline(3); from Lauri Tirkkonen; ok millert@

Use getrusage(2) for the built-in time trial instead of getttimeofday(2).
This makes it possible to measure just the ellapsed user time, which
is what we really care about when benchmarking an algorithm.
OK deraadt@

obvious use for freezero()

In -C mode, process all checksums that match the specified file(s),
not just the first one that matches. OK deraadt@

Move declaration of an 'i' inside the scope of its use. Which in
turn is inside an #ifdef. Thus making the code clearer by eliminating
an #ifdef in the middle of the declarations.

Suggested by millert@

Eliminate some gcc warnings about 'unused variables', mostly by
adding appropriate #ifdef's around declarations.

ok millert@ (with a tweak I will commit separately)

add sha512/256 support to check that the libc code is right :)

Pull in <sys/time.h> for gettimeofday()

all of the update functions take a size_t. correct type and casts.
ok deraadt millert

Remove NULL-check before free(). ok tb@

normalize the ordering of tame requests (particularily, "rpath wpath cpath",
which i have put in that order). this is not important, but helps look
for outliers which might be strange. it hints that "ioctl" should be
reassessed in a few places, to see if "tty" is better; that "unix" may
be used in some places where "route" could now work.

Change all tame callers to namechange to pledge(2).

Repair tame() error check to be == -1

right at startup, this can tame "stdio cpath rpath wpath". after getopt
-h has handled write/creating a file, we can drop to tame "stdio rpath"
since md5 will only read files after that.
i believe i involved lteo for this.

cleanup of the SMALL hiding from Mathias Svensson

Replace <sys/param.h> with <limits.h> and other less dirty headers where
possible. Annotate <sys/param.h> lines with their current reasons. Switch
to PATH_MAX, NGROUPS_MAX, HOST_NAME_MAX+1, LOGIN_NAME_MAX, etc. Change
MIN() and MAX() to local definitions of MINIMUM() and MAXIMUM() where
sensible to avoid pulling in the pollution. These are the files confirmed
through binary verification.
ok guenther, millert, doug (helped with the verification protocol)

Replace all queue *_END macro calls except CIRCLEQ_END with NULL.

CIRCLEQ_* is deprecated and not called in the tree. The other queue types
have *_END macros which were added for symmetry with CIRCLEQ_END. They are
defined as NULL. There's no reason to keep the other *_END macro calls.

ok millert@

Fix memory leak in digest_file() on ferror(). OK tedu@ lteo@

Ignore the -b option if cksum is called as-is (e.g. "cksum -b /bsd")
as documented on the cksum(1) man page.

"looks right" deraadt@
ok millert@

purge sum and sysvsum code. ok deraadt jmc lteo millert
a few initial leftovers spotted by naddy

remove insecure MD4 checksum algorithm; ok tedu@, man page ok jmc@

"A collision attack published in 2007 can find collisions for full
MD4 in less than two hash operations."

When using the -C option, exit with an exit status of 1 if any of the
files specified on the command line do not exist in the checklist.

ok deraadt@ tedu@

add -C to the man page, and adjust usage();
ok lteo

Add support for a -C option. It works on a checklist like -c but only
does the checksum comparison for selected files that are specified on
the command line.

idea discussed with deraadt@ and tedu@
manpage feedback jmc@
feedback/OK deraadt@ millert@

Also move case 'c' into the #ifdef for a smaller binary. It will fall into
default, giving a nice failure. I have not removed -c from the usage()
or getopt() because it is too much butchering...

use -DSHA2_ONLY to be more clear about what we are butchering
idea from tedu

catch up to the fact that md5/sha* got merged, and document -c consistently;
some style and cleanup tweaks while here

ok deraadt

When using a checklist, print MISSING for non-existent files.

Based on an earlier diff by tedu@
Requested by deraadt@
OK deraadt@

the -c option is really a mode change, incompatible with other options.
(there are some others too, but -c is particularly misleading.) split it
out in synopsis and usage.
ok deraadt

Fix wrapping of usage message for cksum.

err() not errx() when fopen fails.

Remove NHASHES, it is not needed and is incompatible with #ifdef SMALL.
Remove style_hash and style_cksum, adjust the style of the selected
algorithms based on rflag and qflag instead.

Remove pmode, fix hfile open mode, check return value when writing
file name to stdout.

Add -DSMALL support to kill a lot of the unncecessary hashes and checksums
that the install media won't need
ok millert, but he used the word butcher!

Add support for -h hashfile. This is very useful with with -p, in a pipeline,
to deliver the hash output to a different location.
ok millert

install a link as sha512. simplify program internals; there are only
two modes. ok deraadt gilles jca

add unsigned char casts for specific calls to ctype.h macros.
ok guenther step

SHA-224 is to SHA-256 as SHA-384 is to SHA-512. It was added in a
later revision of FIPS-180. OK miod@ jmc@ guenther@ djm@

Return an exit code of 1 if the file cannot be opened (e.g. file does
not exist), or if there is an error reading the file stream.

From Patrik Lundin, thanks!

ok deraadt

replace sys/param.h with sys/types.h (and sometimes add limits.h if needed)
ok guenther

fix memset sizeof, found by jsg. ok krw

Use a 32KB data buffer instead of a 1KB buffer.

Use stdio when reading in the file to be checksummed. This can improve
performance due to the internal buffering stdio does wrt the optimal
read size (st_blksize). OK mikeb@

adapt to sha2(3) API changes; ok millert@

md5(1), sha1(1), rmd160(1), cksum(1) and sum(1) have different options;
while here, some KNF improvements.

based on a diff for compress(1) written by millert@,
fixes from guenther@ and pyr@.

ok millert@, pyr@; (for the manual page tweaks) jmc@

properly exit >0 when encountering bad md5 in cheklist

ok gilles@ millert@

From Tim van der Molen <tbvdm.lists at xs4all.nl>

clamp -ttttttttt test counter to 1B, so that it does not wrap negative;
from Nicholas Marriott, ok otto

if we define a constant for the OPTSTRING, we can as well use it.

ok tedu@

allow multiple -t to increase the test count. cpus are getting too fast.
ok grunk. with manpage from grunk and jmc

Fix a cust & pasto
Correctly compute amount of base64 padding in checklist mode

Do not print a warning that no valid checksums were found if there
was a single fingerprint for a file that could not be opened.
The warning that the file could not be opened is sufficient.
From espie@

Supress warnings for the first getopt() pass. Otherwise,
any "unknown option" messages get printed twice.

Add base64 support to cksum and friends. Output encoding can
be set globally or on a per-algorithm basis when using multiple
hash functions. OK espie@ simon@ deraadt@; man help from jmc@

augment the -q option, now if specifed with -c (checklist)
it will only print out the failed cases.

OK millert@

align usage message correctly

simplify synopsis and usage(); ok millert

Add -r and -q options similar to what is supported by FreeBSD's md5
OK deraadt@ ckuethe@

Accept upper and lower case letters in a checksum, since this doesn't
alter the value of the digest. Nor, for that matter, does case affect
the name of the algorithm used.

Partly from Matthew Clarke, clamat (at) telus (dot) net.

ok djm@ millert@

Fix signed/unsigned comparisons by using size_t
Sprinkle some const

ok ray@ jaredy@

Make "cksum -c" grok checklists generated with cksum, sum and sysvsum.
OK deraadt@

Exit unsuccessfully when comparing against a checklist file fails.
OK markus@ millert@

Add support for multiple using algorithms. OK'd by several people
a while ago so get this out of my tree...

GNU md5sum uses tab not space to separate filename and hash. Fixes
parsing of GNU-style checklist files for files with spaces in them.
From Emil Mikulic; closes PR 4027

sync usage w/ man page

Add support for cksum (3 varieties), md4, sha256, sha384, and sha512.
Consequently, cksum moves becomes a link to md5 and moves to /bin.
The cksum man page could use some polishing.

Use the FOO_DIGEST_LENGTH macros

Extend "md5 -c" to parse the output of GNU md5sum.
Adapted from patches by Lawrence Teo and tedu@. OK tedu@

Fix typo: unsigned -> unsigned char; from Thorsten Glaser

fix lots of proto issues

Sync with share/misc/license.template and add missing DARPA credit
where applicable.

ansification, art ok

Use an ISC-tyle license for all my code; it is simpler and more permissive.

return from main

strchr() should be strrchr(); zyrnix

Do not closed stdin; Andrey Matveev

Cast 3 to size_t to shut up stupid gcc on 64bit platforms.

Add -c option similar to GNU md5sum.
Original patch from zyrnix but significantly modified by me.
deraadt@ OK

Report read() errors; noticed and OK by fgs@

Swap TEST_BLOCK_LEN and TEST_BLOCK_COUNT in benchmark printf.

ok millert@

o) __progname aria;

millert@ ok.

Nope, that's not right either. Fix check for mutually exclusive args.

Obvious thinko: command line is wrong if it's too short, not too long.

Fix some typos and add missing test for string of 1 million a's.

New md5/sha1/rmd160 driver with a BSD copyright. Improvements:
o options that should be mutually exclusive are
o time trial works reasonably when it finishes in < 1 sec
o uses a function table instead of a bunch of globals

One single malloc, with a union to get the largest needed size,
should be enough.

use getopt; mph@pobox.com

warn not warnx

Adds rmd160 functionality.

md5(1) is now capable of doing md4, md5 and sha1 digests. Currently
only md5 and sha1 are used.

boost TEST_BLOCK_LEN and TEST_BLOCK_COUNT a bit. this makes '-t' more
useful on fast boxes.
also fflush(stdout) after printing the first test message so
the user knows its actually doing something

64-bit cleanup + pedantic -W flags

-Wall happiness

move to bin/md5 for `greater accessibility'

Use ferror() to check for write error, not the fflush() return value.
We can't rely on buffering to catch write errors in fflush().
Based on a diff from Demi M. Obenour, OK kn@

freezero() is of no use here, the computed digest is not a secret.
ok millert@

I am retiring my old email address; replace it with my OpenBSD one.

fgetln(3) -> getline(3); from Lauri Tirkkonen; ok millert@

Use getrusage(2) for the built-in time trial instead of getttimeofday(2).
This makes it possible to measure just the ellapsed user time, which
is what we really care about when benchmarking an algorithm.
OK deraadt@

obvious use for freezero()

In -C mode, process all checksums that match the specified file(s),
not just the first one that matches. OK deraadt@

Move declaration of an 'i' inside the scope of its use. Which in
turn is inside an #ifdef. Thus making the code clearer by eliminating
an #ifdef in the middle of the declarations.

Suggested by millert@

Eliminate some gcc warnings about 'unused variables', mostly by
adding appropriate #ifdef's around declarations.

ok millert@ (with a tweak I will commit separately)

add sha512/256 support to check that the libc code is right :)

Pull in <sys/time.h> for gettimeofday()

all of the update functions take a size_t. correct type and casts.
ok deraadt millert

Remove NULL-check before free(). ok tb@

normalize the ordering of tame requests (particularily, "rpath wpath cpath",
which i have put in that order). this is not important, but helps look
for outliers which might be strange. it hints that "ioctl" should be
reassessed in a few places, to see if "tty" is better; that "unix" may
be used in some places where "route" could now work.

Change all tame callers to namechange to pledge(2).

Repair tame() error check to be == -1

right at startup, this can tame "stdio cpath rpath wpath". after getopt
-h has handled write/creating a file, we can drop to tame "stdio rpath"
since md5 will only read files after that.
i believe i involved lteo for this.

cleanup of the SMALL hiding from Mathias Svensson

Replace <sys/param.h> with <limits.h> and other less dirty headers where
possible. Annotate <sys/param.h> lines with their current reasons. Switch
to PATH_MAX, NGROUPS_MAX, HOST_NAME_MAX+1, LOGIN_NAME_MAX, etc. Change
MIN() and MAX() to local definitions of MINIMUM() and MAXIMUM() where
sensible to avoid pulling in the pollution. These are the files confirmed
through binary verification.
ok guenther, millert, doug (helped with the verification protocol)

Replace all queue *_END macro calls except CIRCLEQ_END with NULL.

CIRCLEQ_* is deprecated and not called in the tree. The other queue types
have *_END macros which were added for symmetry with CIRCLEQ_END. They are
defined as NULL. There's no reason to keep the other *_END macro calls.

ok millert@

Fix memory leak in digest_file() on ferror(). OK tedu@ lteo@

Ignore the -b option if cksum is called as-is (e.g. "cksum -b /bsd")
as documented on the cksum(1) man page.

"looks right" deraadt@
ok millert@

purge sum and sysvsum code. ok deraadt jmc lteo millert
a few initial leftovers spotted by naddy

remove insecure MD4 checksum algorithm; ok tedu@, man page ok jmc@

"A collision attack published in 2007 can find collisions for full
MD4 in less than two hash operations."

When using the -C option, exit with an exit status of 1 if any of the
files specified on the command line do not exist in the checklist.

ok deraadt@ tedu@

add -C to the man page, and adjust usage();
ok lteo

Add support for a -C option. It works on a checklist like -c but only
does the checksum comparison for selected files that are specified on
the command line.

idea discussed with deraadt@ and tedu@
manpage feedback jmc@
feedback/OK deraadt@ millert@

Also move case 'c' into the #ifdef for a smaller binary. It will fall into
default, giving a nice failure. I have not removed -c from the usage()
or getopt() because it is too much butchering...

use -DSHA2_ONLY to be more clear about what we are butchering
idea from tedu

catch up to the fact that md5/sha* got merged, and document -c consistently;
some style and cleanup tweaks while here

ok deraadt

When using a checklist, print MISSING for non-existent files.

Based on an earlier diff by tedu@
Requested by deraadt@
OK deraadt@

the -c option is really a mode change, incompatible with other options.
(there are some others too, but -c is particularly misleading.) split it
out in synopsis and usage.
ok deraadt

Fix wrapping of usage message for cksum.

err() not errx() when fopen fails.

Remove NHASHES, it is not needed and is incompatible with #ifdef SMALL.
Remove style_hash and style_cksum, adjust the style of the selected
algorithms based on rflag and qflag instead.

Remove pmode, fix hfile open mode, check return value when writing
file name to stdout.

Add -DSMALL support to kill a lot of the unncecessary hashes and checksums
that the install media won't need
ok millert, but he used the word butcher!

Add support for -h hashfile. This is very useful with with -p, in a pipeline,
to deliver the hash output to a different location.
ok millert

install a link as sha512. simplify program internals; there are only
two modes. ok deraadt gilles jca

add unsigned char casts for specific calls to ctype.h macros.
ok guenther step

SHA-224 is to SHA-256 as SHA-384 is to SHA-512. It was added in a
later revision of FIPS-180. OK miod@ jmc@ guenther@ djm@

Return an exit code of 1 if the file cannot be opened (e.g. file does
not exist), or if there is an error reading the file stream.

From Patrik Lundin, thanks!

ok deraadt

replace sys/param.h with sys/types.h (and sometimes add limits.h if needed)
ok guenther

fix memset sizeof, found by jsg. ok krw

Use a 32KB data buffer instead of a 1KB buffer.

Use stdio when reading in the file to be checksummed. This can improve
performance due to the internal buffering stdio does wrt the optimal
read size (st_blksize). OK mikeb@

adapt to sha2(3) API changes; ok millert@

md5(1), sha1(1), rmd160(1), cksum(1) and sum(1) have different options;
while here, some KNF improvements.

based on a diff for compress(1) written by millert@,
fixes from guenther@ and pyr@.

ok millert@, pyr@; (for the manual page tweaks) jmc@

properly exit >0 when encountering bad md5 in cheklist

ok gilles@ millert@

From Tim van der Molen <tbvdm.lists at xs4all.nl>

clamp -ttttttttt test counter to 1B, so that it does not wrap negative;
from Nicholas Marriott, ok otto

if we define a constant for the OPTSTRING, we can as well use it.

ok tedu@

allow multiple -t to increase the test count. cpus are getting too fast.
ok grunk. with manpage from grunk and jmc

Fix a cust & pasto
Correctly compute amount of base64 padding in checklist mode

Do not print a warning that no valid checksums were found if there
was a single fingerprint for a file that could not be opened.
The warning that the file could not be opened is sufficient.
From espie@

Supress warnings for the first getopt() pass. Otherwise,
any "unknown option" messages get printed twice.

Add base64 support to cksum and friends. Output encoding can
be set globally or on a per-algorithm basis when using multiple
hash functions. OK espie@ simon@ deraadt@; man help from jmc@

augment the -q option, now if specifed with -c (checklist)
it will only print out the failed cases.

OK millert@

align usage message correctly

simplify synopsis and usage(); ok millert

Add -r and -q options similar to what is supported by FreeBSD's md5
OK deraadt@ ckuethe@

Accept upper and lower case letters in a checksum, since this doesn't
alter the value of the digest. Nor, for that matter, does case affect
the name of the algorithm used.

Partly from Matthew Clarke, clamat (at) telus (dot) net.

ok djm@ millert@

Fix signed/unsigned comparisons by using size_t
Sprinkle some const

ok ray@ jaredy@

Make "cksum -c" grok checklists generated with cksum, sum and sysvsum.
OK deraadt@

Exit unsuccessfully when comparing against a checklist file fails.
OK markus@ millert@

Add support for multiple using algorithms. OK'd by several people
a while ago so get this out of my tree...

GNU md5sum uses tab not space to separate filename and hash. Fixes
parsing of GNU-style checklist files for files with spaces in them.
From Emil Mikulic; closes PR 4027

sync usage w/ man page

Add support for cksum (3 varieties), md4, sha256, sha384, and sha512.
Consequently, cksum moves becomes a link to md5 and moves to /bin.
The cksum man page could use some polishing.

Use the FOO_DIGEST_LENGTH macros

Extend "md5 -c" to parse the output of GNU md5sum.
Adapted from patches by Lawrence Teo and tedu@. OK tedu@

Fix typo: unsigned -> unsigned char; from Thorsten Glaser

fix lots of proto issues

Sync with share/misc/license.template and add missing DARPA credit
where applicable.

ansification, art ok

Use an ISC-tyle license for all my code; it is simpler and more permissive.

return from main

strchr() should be strrchr(); zyrnix

Do not closed stdin; Andrey Matveev

Cast 3 to size_t to shut up stupid gcc on 64bit platforms.

Add -c option similar to GNU md5sum.
Original patch from zyrnix but significantly modified by me.
deraadt@ OK

Report read() errors; noticed and OK by fgs@

Swap TEST_BLOCK_LEN and TEST_BLOCK_COUNT in benchmark printf.

ok millert@

o) __progname aria;

millert@ ok.

Nope, that's not right either. Fix check for mutually exclusive args.

Obvious thinko: command line is wrong if it's too short, not too long.

Fix some typos and add missing test for string of 1 million a's.

New md5/sha1/rmd160 driver with a BSD copyright. Improvements:
o options that should be mutually exclusive are
o time trial works reasonably when it finishes in < 1 sec
o uses a function table instead of a bunch of globals

One single malloc, with a union to get the largest needed size,
should be enough.

use getopt; mph@pobox.com

warn not warnx

Adds rmd160 functionality.

md5(1) is now capable of doing md4, md5 and sha1 digests. Currently
only md5 and sha1 are used.

boost TEST_BLOCK_LEN and TEST_BLOCK_COUNT a bit. this makes '-t' more
useful on fast boxes.
also fflush(stdout) after printing the first test message so
the user knows its actually doing something

64-bit cleanup + pedantic -W flags

-Wall happiness

move to bin/md5 for `greater accessibility'

freezero() is of no use here, the computed digest is not a secret.
ok millert@

I am retiring my old email address; replace it with my OpenBSD one.

fgetln(3) -> getline(3); from Lauri Tirkkonen; ok millert@

Use getrusage(2) for the built-in time trial instead of getttimeofday(2).
This makes it possible to measure just the ellapsed user time, which
is what we really care about when benchmarking an algorithm.
OK deraadt@

obvious use for freezero()

In -C mode, process all checksums that match the specified file(s),
not just the first one that matches. OK deraadt@

Move declaration of an 'i' inside the scope of its use. Which in
turn is inside an #ifdef. Thus making the code clearer by eliminating
an #ifdef in the middle of the declarations.

Suggested by millert@

Eliminate some gcc warnings about 'unused variables', mostly by
adding appropriate #ifdef's around declarations.

ok millert@ (with a tweak I will commit separately)

add sha512/256 support to check that the libc code is right :)

Pull in <sys/time.h> for gettimeofday()

all of the update functions take a size_t. correct type and casts.
ok deraadt millert

Remove NULL-check before free(). ok tb@

normalize the ordering of tame requests (particularily, "rpath wpath cpath",
which i have put in that order). this is not important, but helps look
for outliers which might be strange. it hints that "ioctl" should be
reassessed in a few places, to see if "tty" is better; that "unix" may
be used in some places where "route" could now work.

Change all tame callers to namechange to pledge(2).

Repair tame() error check to be == -1

right at startup, this can tame "stdio cpath rpath wpath". after getopt
-h has handled write/creating a file, we can drop to tame "stdio rpath"
since md5 will only read files after that.
i believe i involved lteo for this.

cleanup of the SMALL hiding from Mathias Svensson

Replace <sys/param.h> with <limits.h> and other less dirty headers where
possible. Annotate <sys/param.h> lines with their current reasons. Switch
to PATH_MAX, NGROUPS_MAX, HOST_NAME_MAX+1, LOGIN_NAME_MAX, etc. Change
MIN() and MAX() to local definitions of MINIMUM() and MAXIMUM() where
sensible to avoid pulling in the pollution. These are the files confirmed
through binary verification.
ok guenther, millert, doug (helped with the verification protocol)

Replace all queue *_END macro calls except CIRCLEQ_END with NULL.

CIRCLEQ_* is deprecated and not called in the tree. The other queue types
have *_END macros which were added for symmetry with CIRCLEQ_END. They are
defined as NULL. There's no reason to keep the other *_END macro calls.

ok millert@

Fix memory leak in digest_file() on ferror(). OK tedu@ lteo@

Ignore the -b option if cksum is called as-is (e.g. "cksum -b /bsd")
as documented on the cksum(1) man page.

"looks right" deraadt@
ok millert@

purge sum and sysvsum code. ok deraadt jmc lteo millert
a few initial leftovers spotted by naddy

remove insecure MD4 checksum algorithm; ok tedu@, man page ok jmc@

"A collision attack published in 2007 can find collisions for full
MD4 in less than two hash operations."

When using the -C option, exit with an exit status of 1 if any of the
files specified on the command line do not exist in the checklist.

ok deraadt@ tedu@

add -C to the man page, and adjust usage();
ok lteo

Add support for a -C option. It works on a checklist like -c but only
does the checksum comparison for selected files that are specified on
the command line.

idea discussed with deraadt@ and tedu@
manpage feedback jmc@
feedback/OK deraadt@ millert@

Also move case 'c' into the #ifdef for a smaller binary. It will fall into
default, giving a nice failure. I have not removed -c from the usage()
or getopt() because it is too much butchering...

use -DSHA2_ONLY to be more clear about what we are butchering
idea from tedu

catch up to the fact that md5/sha* got merged, and document -c consistently;
some style and cleanup tweaks while here

ok deraadt

When using a checklist, print MISSING for non-existent files.

Based on an earlier diff by tedu@
Requested by deraadt@
OK deraadt@

the -c option is really a mode change, incompatible with other options.
(there are some others too, but -c is particularly misleading.) split it
out in synopsis and usage.
ok deraadt

Fix wrapping of usage message for cksum.

err() not errx() when fopen fails.

Remove NHASHES, it is not needed and is incompatible with #ifdef SMALL.
Remove style_hash and style_cksum, adjust the style of the selected
algorithms based on rflag and qflag instead.

Remove pmode, fix hfile open mode, check return value when writing
file name to stdout.

Add -DSMALL support to kill a lot of the unncecessary hashes and checksums
that the install media won't need
ok millert, but he used the word butcher!

Add support for -h hashfile. This is very useful with with -p, in a pipeline,
to deliver the hash output to a different location.
ok millert

install a link as sha512. simplify program internals; there are only
two modes. ok deraadt gilles jca

add unsigned char casts for specific calls to ctype.h macros.
ok guenther step

SHA-224 is to SHA-256 as SHA-384 is to SHA-512. It was added in a
later revision of FIPS-180. OK miod@ jmc@ guenther@ djm@

Return an exit code of 1 if the file cannot be opened (e.g. file does
not exist), or if there is an error reading the file stream.

From Patrik Lundin, thanks!

ok deraadt

replace sys/param.h with sys/types.h (and sometimes add limits.h if needed)
ok guenther

fix memset sizeof, found by jsg. ok krw

Use a 32KB data buffer instead of a 1KB buffer.

Use stdio when reading in the file to be checksummed. This can improve
performance due to the internal buffering stdio does wrt the optimal
read size (st_blksize). OK mikeb@

adapt to sha2(3) API changes; ok millert@

md5(1), sha1(1), rmd160(1), cksum(1) and sum(1) have different options;
while here, some KNF improvements.

based on a diff for compress(1) written by millert@,
fixes from guenther@ and pyr@.

ok millert@, pyr@; (for the manual page tweaks) jmc@

properly exit >0 when encountering bad md5 in cheklist

ok gilles@ millert@

From Tim van der Molen <tbvdm.lists at xs4all.nl>

clamp -ttttttttt test counter to 1B, so that it does not wrap negative;
from Nicholas Marriott, ok otto

if we define a constant for the OPTSTRING, we can as well use it.

ok tedu@

allow multiple -t to increase the test count. cpus are getting too fast.
ok grunk. with manpage from grunk and jmc

Fix a cust & pasto
Correctly compute amount of base64 padding in checklist mode

Do not print a warning that no valid checksums were found if there
was a single fingerprint for a file that could not be opened.
The warning that the file could not be opened is sufficient.
From espie@

Supress warnings for the first getopt() pass. Otherwise,
any "unknown option" messages get printed twice.

Add base64 support to cksum and friends. Output encoding can
be set globally or on a per-algorithm basis when using multiple
hash functions. OK espie@ simon@ deraadt@; man help from jmc@

augment the -q option, now if specifed with -c (checklist)
it will only print out the failed cases.

OK millert@

align usage message correctly

simplify synopsis and usage(); ok millert

Add -r and -q options similar to what is supported by FreeBSD's md5
OK deraadt@ ckuethe@

Accept upper and lower case letters in a checksum, since this doesn't
alter the value of the digest. Nor, for that matter, does case affect
the name of the algorithm used.

Partly from Matthew Clarke, clamat (at) telus (dot) net.

ok djm@ millert@

Fix signed/unsigned comparisons by using size_t
Sprinkle some const

ok ray@ jaredy@

Make "cksum -c" grok checklists generated with cksum, sum and sysvsum.
OK deraadt@

Exit unsuccessfully when comparing against a checklist file fails.
OK markus@ millert@

Add support for multiple using algorithms. OK'd by several people
a while ago so get this out of my tree...

GNU md5sum uses tab not space to separate filename and hash. Fixes
parsing of GNU-style checklist files for files with spaces in them.
From Emil Mikulic; closes PR 4027

sync usage w/ man page

Add support for cksum (3 varieties), md4, sha256, sha384, and sha512.
Consequently, cksum moves becomes a link to md5 and moves to /bin.
The cksum man page could use some polishing.

Use the FOO_DIGEST_LENGTH macros

Extend "md5 -c" to parse the output of GNU md5sum.
Adapted from patches by Lawrence Teo and tedu@. OK tedu@

Fix typo: unsigned -> unsigned char; from Thorsten Glaser

fix lots of proto issues

Sync with share/misc/license.template and add missing DARPA credit
where applicable.

ansification, art ok

Use an ISC-tyle license for all my code; it is simpler and more permissive.

return from main

strchr() should be strrchr(); zyrnix

Do not closed stdin; Andrey Matveev

Cast 3 to size_t to shut up stupid gcc on 64bit platforms.

Add -c option similar to GNU md5sum.
Original patch from zyrnix but significantly modified by me.
deraadt@ OK

Report read() errors; noticed and OK by fgs@

Swap TEST_BLOCK_LEN and TEST_BLOCK_COUNT in benchmark printf.

ok millert@

o) __progname aria;

millert@ ok.

Nope, that's not right either. Fix check for mutually exclusive args.

Obvious thinko: command line is wrong if it's too short, not too long.

Fix some typos and add missing test for string of 1 million a's.

New md5/sha1/rmd160 driver with a BSD copyright. Improvements:
o options that should be mutually exclusive are
o time trial works reasonably when it finishes in < 1 sec
o uses a function table instead of a bunch of globals

One single malloc, with a union to get the largest needed size,
should be enough.

use getopt; mph@pobox.com

warn not warnx

Adds rmd160 functionality.

md5(1) is now capable of doing md4, md5 and sha1 digests. Currently
only md5 and sha1 are used.

boost TEST_BLOCK_LEN and TEST_BLOCK_COUNT a bit. this makes '-t' more
useful on fast boxes.
also fflush(stdout) after printing the first test message so
the user knows its actually doing something

64-bit cleanup + pedantic -W flags

-Wall happiness

move to bin/md5 for `greater accessibility'

I am retiring my old email address; replace it with my OpenBSD one.

fgetln(3) -> getline(3); from Lauri Tirkkonen; ok millert@

Use getrusage(2) for the built-in time trial instead of getttimeofday(2).
This makes it possible to measure just the ellapsed user time, which
is what we really care about when benchmarking an algorithm.
OK deraadt@

obvious use for freezero()

In -C mode, process all checksums that match the specified file(s),
not just the first one that matches. OK deraadt@

Move declaration of an 'i' inside the scope of its use. Which in
turn is inside an #ifdef. Thus making the code clearer by eliminating
an #ifdef in the middle of the declarations.

Suggested by millert@

Eliminate some gcc warnings about 'unused variables', mostly by
adding appropriate #ifdef's around declarations.

ok millert@ (with a tweak I will commit separately)

add sha512/256 support to check that the libc code is right :)

Pull in <sys/time.h> for gettimeofday()

all of the update functions take a size_t. correct type and casts.
ok deraadt millert

Remove NULL-check before free(). ok tb@

normalize the ordering of tame requests (particularily, "rpath wpath cpath",
which i have put in that order). this is not important, but helps look
for outliers which might be strange. it hints that "ioctl" should be
reassessed in a few places, to see if "tty" is better; that "unix" may
be used in some places where "route" could now work.

Change all tame callers to namechange to pledge(2).

Repair tame() error check to be == -1

right at startup, this can tame "stdio cpath rpath wpath". after getopt
-h has handled write/creating a file, we can drop to tame "stdio rpath"
since md5 will only read files after that.
i believe i involved lteo for this.

cleanup of the SMALL hiding from Mathias Svensson

Replace <sys/param.h> with <limits.h> and other less dirty headers where
possible. Annotate <sys/param.h> lines with their current reasons. Switch
to PATH_MAX, NGROUPS_MAX, HOST_NAME_MAX+1, LOGIN_NAME_MAX, etc. Change
MIN() and MAX() to local definitions of MINIMUM() and MAXIMUM() where
sensible to avoid pulling in the pollution. These are the files confirmed
through binary verification.
ok guenther, millert, doug (helped with the verification protocol)

Replace all queue *_END macro calls except CIRCLEQ_END with NULL.

CIRCLEQ_* is deprecated and not called in the tree. The other queue types
have *_END macros which were added for symmetry with CIRCLEQ_END. They are
defined as NULL. There's no reason to keep the other *_END macro calls.

ok millert@

Fix memory leak in digest_file() on ferror(). OK tedu@ lteo@

Ignore the -b option if cksum is called as-is (e.g. "cksum -b /bsd")
as documented on the cksum(1) man page.

"looks right" deraadt@
ok millert@

purge sum and sysvsum code. ok deraadt jmc lteo millert
a few initial leftovers spotted by naddy

remove insecure MD4 checksum algorithm; ok tedu@, man page ok jmc@

"A collision attack published in 2007 can find collisions for full
MD4 in less than two hash operations."

When using the -C option, exit with an exit status of 1 if any of the
files specified on the command line do not exist in the checklist.

ok deraadt@ tedu@

add -C to the man page, and adjust usage();
ok lteo

Add support for a -C option. It works on a checklist like -c but only
does the checksum comparison for selected files that are specified on
the command line.

idea discussed with deraadt@ and tedu@
manpage feedback jmc@
feedback/OK deraadt@ millert@

Also move case 'c' into the #ifdef for a smaller binary. It will fall into
default, giving a nice failure. I have not removed -c from the usage()
or getopt() because it is too much butchering...

use -DSHA2_ONLY to be more clear about what we are butchering
idea from tedu

catch up to the fact that md5/sha* got merged, and document -c consistently;
some style and cleanup tweaks while here

ok deraadt

When using a checklist, print MISSING for non-existent files.

Based on an earlier diff by tedu@
Requested by deraadt@
OK deraadt@

the -c option is really a mode change, incompatible with other options.
(there are some others too, but -c is particularly misleading.) split it
out in synopsis and usage.
ok deraadt

Fix wrapping of usage message for cksum.

err() not errx() when fopen fails.

Remove NHASHES, it is not needed and is incompatible with #ifdef SMALL.
Remove style_hash and style_cksum, adjust the style of the selected
algorithms based on rflag and qflag instead.

Remove pmode, fix hfile open mode, check return value when writing
file name to stdout.

Add -DSMALL support to kill a lot of the unncecessary hashes and checksums
that the install media won't need
ok millert, but he used the word butcher!

Add support for -h hashfile. This is very useful with with -p, in a pipeline,
to deliver the hash output to a different location.
ok millert

install a link as sha512. simplify program internals; there are only
two modes. ok deraadt gilles jca

add unsigned char casts for specific calls to ctype.h macros.
ok guenther step

SHA-224 is to SHA-256 as SHA-384 is to SHA-512. It was added in a
later revision of FIPS-180. OK miod@ jmc@ guenther@ djm@

Return an exit code of 1 if the file cannot be opened (e.g. file does
not exist), or if there is an error reading the file stream.

From Patrik Lundin, thanks!

ok deraadt

replace sys/param.h with sys/types.h (and sometimes add limits.h if needed)
ok guenther

fix memset sizeof, found by jsg. ok krw

Use a 32KB data buffer instead of a 1KB buffer.

Use stdio when reading in the file to be checksummed. This can improve
performance due to the internal buffering stdio does wrt the optimal
read size (st_blksize). OK mikeb@

adapt to sha2(3) API changes; ok millert@

md5(1), sha1(1), rmd160(1), cksum(1) and sum(1) have different options;
while here, some KNF improvements.

based on a diff for compress(1) written by millert@,
fixes from guenther@ and pyr@.

ok millert@, pyr@; (for the manual page tweaks) jmc@

properly exit >0 when encountering bad md5 in cheklist

ok gilles@ millert@

From Tim van der Molen <tbvdm.lists at xs4all.nl>

clamp -ttttttttt test counter to 1B, so that it does not wrap negative;
from Nicholas Marriott, ok otto

if we define a constant for the OPTSTRING, we can as well use it.

ok tedu@

allow multiple -t to increase the test count. cpus are getting too fast.
ok grunk. with manpage from grunk and jmc

Fix a cust & pasto
Correctly compute amount of base64 padding in checklist mode

Do not print a warning that no valid checksums were found if there
was a single fingerprint for a file that could not be opened.
The warning that the file could not be opened is sufficient.
From espie@

Supress warnings for the first getopt() pass. Otherwise,
any "unknown option" messages get printed twice.

Add base64 support to cksum and friends. Output encoding can
be set globally or on a per-algorithm basis when using multiple
hash functions. OK espie@ simon@ deraadt@; man help from jmc@

augment the -q option, now if specifed with -c (checklist)
it will only print out the failed cases.

OK millert@

align usage message correctly

simplify synopsis and usage(); ok millert

Add -r and -q options similar to what is supported by FreeBSD's md5
OK deraadt@ ckuethe@

Accept upper and lower case letters in a checksum, since this doesn't
alter the value of the digest. Nor, for that matter, does case affect
the name of the algorithm used.

Partly from Matthew Clarke, clamat (at) telus (dot) net.

ok djm@ millert@