pax: don't overwrite destination if -r -w copy fails

Add more error handling to pax -r -w so that any failure
during the copy to the temporary file (including a failed flush)
prevents any existing destination file from being replaced
with the partial (including possibly empty) temporary file.
The partial temporary file is removed. pax still exists non-zero.

Thanks to Michael van Elst (mlelstv@) for the analysis
of the problem in the PR.

Should fix PR misc/33753.

pax: fix typos in comments in file_subs.c & tar.c

Stamp out "greengrocers' apostrophes" in various places (arguably there
are still more present, but style guides vary on that, and my energies
spent corralling wayward punctuation marks could be spent elsewhere).

branches: 1.63.26;
PR/48095: NAKAJIMA Yoshihiro: remove casts from time_t to long

branches: 1.62.6; 1.62.12;
Fixes from christos for pr-41167

XXX -X is still broken, but that wasn't part of this PR.

branches: 1.61.10; 1.61.12;
Teach set_ftime about symbolic links, because it has to know: on some
Linux systems, when we build as a tool we think we have lutimes but it
doesn't work on some filesystems at runtime. A bit ugly but effective
and without use of AC_TRY_RUN in the tool build. Tidier (than mine)
set_ftime reorganization from christos.

branches: 1.60.4;
fix typos

Change all the 'return(x)' to 'return c'.
Makes some other searches I want to do a bit easier.

obey umask as it used to be before the "mkstemp and then rename" change.

Check for another race reported by chuq.

Another missing mkdir() race pointed out by chuq. Factor out the mkdir code.

use stat and not lstat to detect if we have a directory. refactor the mkpath()
code.

when creating directories, check if the failure occured because someone
else created the directory before we did to avoid races. From chuq.

For clarity instead of using -1 and -2 use the negative of the archive type.

Keep track of the length of the gnu long name/link hack, so that
we can append to it. Modify the code so that file_write is re-entrant,
even in the gnu long name/link hack.

The old code assumed that the buffer already contained the necessary
blocks to satisfy the read file request that contained the long
filename. This is not always the case, specially when we are dealing
with pipes which do shorter reads, thus having more probability
that a long file name will fall across a buffer boundary.

To reproduce, create a tar archive with a lot of gnu-long-names
(pkgsrc/devel/libsigc++2 is a good example), do a tar -tf to get
a list of filenames, compress it and do a tar -tzf to get another
list of the filenames. Notice that the two lists differ.

Propery handle "cpio" archives where the last hardlink includes the
data of a file. This fixes PR bin/26514.

Salvage the `create parent directories and try again' path which is
removed by the previous commit.

break out of the loop and avoid the infinite loop if attempts to
create, unlink and create all fail.

don't attempt to remove the current directory.

Completely rework how tools/compat is done. Purge all uses/references to
_NETBSD_SOURCE as this makes cross building from older/newer versions of
NetBSD harder, not easier (and also makes the resulting tools 'different')

Wrap all required code with the inclusion of nbtool_config.h, attempt to
only use POSIX code in all places (or when reasonable test w. configure and
provide definitions: ala u_int, etc).

Reviewed by lukem. Tested on FreeBSD 4.9, Redhat Linux ES3, NetBSD 1.6.2 x86
NetBSD current (x86 and amd64) and Solaris 9.

Fixes PR's: PR#17762 PR#25944

Don't exit with 1 on broken archives that mark directories as files for
the benefit of pkgsrc. I would have prefered that people fixed their
archivers instead.

Set the chflags *after* the rename, not before.

PR/25350: Alan Barrett: Treat the sticky bit specially as non-superusers
cannot set it on non-directories (elvis-2.2_0 extraction failed)

PR/25188: Thomas Klausner: pax-as-tar problem with trailing slashes
The problem is with the program that generates the tar file:
-rwxrwxrwx 1 root wheel 0 Feb 8 16:46 faad2/aacDECdrop/
It creates directory nodes without the 'd' bit set, so that pax thinks
they are files and does the temporary name and dance with them. Added
code to detect this condition, warn about it, and work around it.

PR/25207: Hubert Feyrer: tar cores on certain archive contents
Don't forget to cleanup xtmp_name when we fail to create the temporary file.

branches: 1.41.2;
When pax catches a signal and cleans up, make sure to remove any temporary
file that was being extracted to. This will prevent pax from leaving
droppings when you hit ^C.

Nuke extra mkstemp. Use name, not tmp_name with chk_path.

Back out last change. Since the name is the entire path, prefixing it
with a '.' will not have the desired effect.

When creating the temporary file, prefix the name with a '.'

When restoring files from an archive, initially restore to a temp-file and
once it has been instantiated correctly, rename it to desired name. This
prevents the problem of partially created files being accessed before they
are complete. If said file is a shared library, that can cause ramdon core
dumps.

Overhaul how `build.sh tools' are used:

* Rename "config.h" to "nbtool_config.h" and
HAVE_CONFIG_H to HAVE_NBTOOL_CONFIG_H.
This makes in more obvious in the source when we're using
tools/compat/config.h versus "standard autoconf" config.h

* Consistently move the inclusion of nbtool_config.h to before
<sys/cdefs.h> so that the former can provide __RCSID() (et al),
and there's no need to protect those macros any more.

These changes should make it easier to "tool-ify" a program by adding:
#if HAVE_NBTOOL_CONFIG_H
#include "nbtool_config.h"
#endif
to the top of the source files (for the general case).

Move Keith Muller's code from a 4-clause to a 3-clause licence by removing
the advertising clause. Diffs provided in PR 22397 by Joel Baker, confirmed
to the board by Keith Muller.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22249, verified by myself.

no need for HAVE_LCHOWN and HAVE_LCHMOD anymore.

add missing braces.

from HIRATSUKA Kouichirou in PR bin/20276.

cosmetic error condition improvements.
- sprinkle some fflush() to print filenames correctly.
- print a \n to stdout to fix formatting.
- English, consistently use "Cannot" in error messages.

some whitespace cleanup.

Fix various spelling, grammar, and punctuation problems in comments, from Sergey Svishchev in PR 20164.

The Double-Semi-Colon Police.

writable, not writeable.

Fix oops in last, chown() failure should be fatal.

Noted by Charles Hannum.

Don't attempt to chown files if euid != 0, and make chown() non-fatal
if it does fail.

This addresses my PR bin/18912 (running pkg_info as non-root user on a
binary package).

Add proper longlink support. Previously we handled longname support, and
the longlink support was completely wrong.

merge OpenBSD changes:
- correct -C processing
- add ability to read filenames and flags from a file
- don't print dangerous escape sequences to the terminal
- use strlcpy/strncpy properly.
- handle tmpfile creation better.
- improve documentation of options.
- handle stdout/stderr list selection correctly.
- kill gzip when we get interrupted.
- simplify gzip setup.
- add more flags to programs.

additional changes:
- librmt processing.
- set POSIXLY_CORRECT in options parsing.
- prevent more string overruns.
- support -T

we don't turn the switch on to replace tar and cpio yet.

branches: 1.23.2;
"doesn't" should have an 's'.

Protect __RCSID and __COPYRIGHT from being invoked if not defined.

Make almost all tools compile and run properly on non-NetBSD hosts. (In
particular, most tools now run correctly on Solaris 7.)

- ansi KNF (just remove all the #ifndef __STDC__ prototype junk,
leaving the ansi stuff)
- use longlong_t instead of quad_t (etc), and rename *uqd*() -> *ull*()
- clean up the NET2_STAT stuff similar to ftpd; provide #defines and
macros which select which cast to use, etc
- clean up the NET2_FTS and NET2_REGEX #define use

minor KNF nits.

branches: 1.18.2;
Fix doubled 'the' in comment.

Cleanup and fix typos.
Partially from FreeBSD.

Fix SIGINFO botch (PR #8868).
Continue partial write(2) on signals (xwrite()).
Partial read(2) at a few places are also continued (xread()).

Add {read,write}_with_restart() hooks for porting on systems
which don't restart interrupted read()/write() calls.

Reviewed and discussed in tech-kern and tech-userlevel lists.

branches: 1.15.2;
Completely disable all the chflags(2) support for now. With this enabled, I
can't even get a build to finish over NFS. Not acceptable.

Don't try to set file flags for a symlink.

add a new -p subflag: 'f' to preserve 4.4BSD fileflags. this only
really works for `pax -rw' but that's where it is most useful.

support ././@LongLink extraction, as created by GNU tar.

remove some old cruft.

branches: 1.10.4;
fix typo in a comment

Deal with GCC warnings.

When copying time stamps, copy the microseconds, too. Also nuke silly
static structure.

Restore owner/group/mode/atime/mtime of symbolic links.

- Rename local warn so tty_warn that we can include <err.h> since this program
uses err() and errx().
- Fix printf format strings.
- Added WARNS=1; note
- Added missing prototypes.
- Split overloaded trail function into trail() and subtrail().

kill 'register'

convert to new RCS id conventions.

Fix up RCS ids.

Add RCS ID's

branches: 1.1.1;
Initial revision

pax: fix typos in comments in file_subs.c & tar.c

Stamp out "greengrocers' apostrophes" in various places (arguably there
are still more present, but style guides vary on that, and my energies
spent corralling wayward punctuation marks could be spent elsewhere).

PR/48095: NAKAJIMA Yoshihiro: remove casts from time_t to long

branches: 1.62.6; 1.62.12;
Fixes from christos for pr-41167

XXX -X is still broken, but that wasn't part of this PR.

branches: 1.61.10; 1.61.12;
Teach set_ftime about symbolic links, because it has to know: on some
Linux systems, when we build as a tool we think we have lutimes but it
doesn't work on some filesystems at runtime. A bit ugly but effective
and without use of AC_TRY_RUN in the tool build. Tidier (than mine)
set_ftime reorganization from christos.

branches: 1.60.4;
fix typos

Change all the 'return(x)' to 'return c'.
Makes some other searches I want to do a bit easier.

obey umask as it used to be before the "mkstemp and then rename" change.

Check for another race reported by chuq.

Another missing mkdir() race pointed out by chuq. Factor out the mkdir code.

use stat and not lstat to detect if we have a directory. refactor the mkpath()
code.

when creating directories, check if the failure occured because someone
else created the directory before we did to avoid races. From chuq.

For clarity instead of using -1 and -2 use the negative of the archive type.

Keep track of the length of the gnu long name/link hack, so that
we can append to it. Modify the code so that file_write is re-entrant,
even in the gnu long name/link hack.

The old code assumed that the buffer already contained the necessary
blocks to satisfy the read file request that contained the long
filename. This is not always the case, specially when we are dealing
with pipes which do shorter reads, thus having more probability
that a long file name will fall across a buffer boundary.

To reproduce, create a tar archive with a lot of gnu-long-names
(pkgsrc/devel/libsigc++2 is a good example), do a tar -tf to get
a list of filenames, compress it and do a tar -tzf to get another
list of the filenames. Notice that the two lists differ.

Propery handle "cpio" archives where the last hardlink includes the
data of a file. This fixes PR bin/26514.

Salvage the `create parent directories and try again' path which is
removed by the previous commit.

break out of the loop and avoid the infinite loop if attempts to
create, unlink and create all fail.

don't attempt to remove the current directory.

Completely rework how tools/compat is done. Purge all uses/references to
_NETBSD_SOURCE as this makes cross building from older/newer versions of
NetBSD harder, not easier (and also makes the resulting tools 'different')

Wrap all required code with the inclusion of nbtool_config.h, attempt to
only use POSIX code in all places (or when reasonable test w. configure and
provide definitions: ala u_int, etc).

Reviewed by lukem. Tested on FreeBSD 4.9, Redhat Linux ES3, NetBSD 1.6.2 x86
NetBSD current (x86 and amd64) and Solaris 9.

Fixes PR's: PR#17762 PR#25944

Don't exit with 1 on broken archives that mark directories as files for
the benefit of pkgsrc. I would have prefered that people fixed their
archivers instead.

Set the chflags *after* the rename, not before.

PR/25350: Alan Barrett: Treat the sticky bit specially as non-superusers
cannot set it on non-directories (elvis-2.2_0 extraction failed)

PR/25188: Thomas Klausner: pax-as-tar problem with trailing slashes
The problem is with the program that generates the tar file:
-rwxrwxrwx 1 root wheel 0 Feb 8 16:46 faad2/aacDECdrop/
It creates directory nodes without the 'd' bit set, so that pax thinks
they are files and does the temporary name and dance with them. Added
code to detect this condition, warn about it, and work around it.

PR/25207: Hubert Feyrer: tar cores on certain archive contents
Don't forget to cleanup xtmp_name when we fail to create the temporary file.

branches: 1.41.2;
When pax catches a signal and cleans up, make sure to remove any temporary
file that was being extracted to. This will prevent pax from leaving
droppings when you hit ^C.

Nuke extra mkstemp. Use name, not tmp_name with chk_path.

Back out last change. Since the name is the entire path, prefixing it
with a '.' will not have the desired effect.

When creating the temporary file, prefix the name with a '.'

When restoring files from an archive, initially restore to a temp-file and
once it has been instantiated correctly, rename it to desired name. This
prevents the problem of partially created files being accessed before they
are complete. If said file is a shared library, that can cause ramdon core
dumps.

Overhaul how `build.sh tools' are used:

* Rename "config.h" to "nbtool_config.h" and
HAVE_CONFIG_H to HAVE_NBTOOL_CONFIG_H.
This makes in more obvious in the source when we're using
tools/compat/config.h versus "standard autoconf" config.h

* Consistently move the inclusion of nbtool_config.h to before
<sys/cdefs.h> so that the former can provide __RCSID() (et al),
and there's no need to protect those macros any more.

These changes should make it easier to "tool-ify" a program by adding:
#if HAVE_NBTOOL_CONFIG_H
#include "nbtool_config.h"
#endif
to the top of the source files (for the general case).

Move Keith Muller's code from a 4-clause to a 3-clause licence by removing
the advertising clause. Diffs provided in PR 22397 by Joel Baker, confirmed
to the board by Keith Muller.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22249, verified by myself.

no need for HAVE_LCHOWN and HAVE_LCHMOD anymore.

add missing braces.

from HIRATSUKA Kouichirou in PR bin/20276.

cosmetic error condition improvements.
- sprinkle some fflush() to print filenames correctly.
- print a \n to stdout to fix formatting.
- English, consistently use "Cannot" in error messages.

some whitespace cleanup.

Fix various spelling, grammar, and punctuation problems in comments, from Sergey Svishchev in PR 20164.

The Double-Semi-Colon Police.

writable, not writeable.

Fix oops in last, chown() failure should be fatal.

Noted by Charles Hannum.

Don't attempt to chown files if euid != 0, and make chown() non-fatal
if it does fail.

This addresses my PR bin/18912 (running pkg_info as non-root user on a
binary package).

Add proper longlink support. Previously we handled longname support, and
the longlink support was completely wrong.

merge OpenBSD changes:
- correct -C processing
- add ability to read filenames and flags from a file
- don't print dangerous escape sequences to the terminal
- use strlcpy/strncpy properly.
- handle tmpfile creation better.
- improve documentation of options.
- handle stdout/stderr list selection correctly.
- kill gzip when we get interrupted.
- simplify gzip setup.
- add more flags to programs.

additional changes:
- librmt processing.
- set POSIXLY_CORRECT in options parsing.
- prevent more string overruns.
- support -T

we don't turn the switch on to replace tar and cpio yet.

branches: 1.23.2;
"doesn't" should have an 's'.

Protect __RCSID and __COPYRIGHT from being invoked if not defined.

Make almost all tools compile and run properly on non-NetBSD hosts. (In
particular, most tools now run correctly on Solaris 7.)

- ansi KNF (just remove all the #ifndef __STDC__ prototype junk,
leaving the ansi stuff)
- use longlong_t instead of quad_t (etc), and rename *uqd*() -> *ull*()
- clean up the NET2_STAT stuff similar to ftpd; provide #defines and
macros which select which cast to use, etc
- clean up the NET2_FTS and NET2_REGEX #define use

minor KNF nits.

branches: 1.18.2;
Fix doubled 'the' in comment.

Cleanup and fix typos.
Partially from FreeBSD.

Fix SIGINFO botch (PR #8868).
Continue partial write(2) on signals (xwrite()).
Partial read(2) at a few places are also continued (xread()).

Add {read,write}_with_restart() hooks for porting on systems
which don't restart interrupted read()/write() calls.

Reviewed and discussed in tech-kern and tech-userlevel lists.

branches: 1.15.2;
Completely disable all the chflags(2) support for now. With this enabled, I
can't even get a build to finish over NFS. Not acceptable.

Don't try to set file flags for a symlink.

add a new -p subflag: 'f' to preserve 4.4BSD fileflags. this only
really works for `pax -rw' but that's where it is most useful.

support ././@LongLink extraction, as created by GNU tar.

remove some old cruft.

branches: 1.10.4;
fix typo in a comment

Deal with GCC warnings.

When copying time stamps, copy the microseconds, too. Also nuke silly
static structure.

Restore owner/group/mode/atime/mtime of symbolic links.

- Rename local warn so tty_warn that we can include <err.h> since this program
uses err() and errx().
- Fix printf format strings.
- Added WARNS=1; note
- Added missing prototypes.
- Split overloaded trail function into trail() and subtrail().

kill 'register'

convert to new RCS id conventions.

Fix up RCS ids.

Add RCS ID's

branches: 1.1.1;
Initial revision

PR/48095: NAKAJIMA Yoshihiro: remove casts from time_t to long

branches: 1.62.6; 1.62.12;
Fixes from christos for pr-41167

XXX -X is still broken, but that wasn't part of this PR.

branches: 1.61.10; 1.61.12;
Teach set_ftime about symbolic links, because it has to know: on some
Linux systems, when we build as a tool we think we have lutimes but it
doesn't work on some filesystems at runtime. A bit ugly but effective
and without use of AC_TRY_RUN in the tool build. Tidier (than mine)
set_ftime reorganization from christos.

branches: 1.60.4;
fix typos

Change all the 'return(x)' to 'return c'.
Makes some other searches I want to do a bit easier.

obey umask as it used to be before the "mkstemp and then rename" change.

Check for another race reported by chuq.

Another missing mkdir() race pointed out by chuq. Factor out the mkdir code.

use stat and not lstat to detect if we have a directory. refactor the mkpath()
code.

when creating directories, check if the failure occured because someone
else created the directory before we did to avoid races. From chuq.

For clarity instead of using -1 and -2 use the negative of the archive type.

Keep track of the length of the gnu long name/link hack, so that
we can append to it. Modify the code so that file_write is re-entrant,
even in the gnu long name/link hack.

The old code assumed that the buffer already contained the necessary
blocks to satisfy the read file request that contained the long
filename. This is not always the case, specially when we are dealing
with pipes which do shorter reads, thus having more probability
that a long file name will fall across a buffer boundary.

To reproduce, create a tar archive with a lot of gnu-long-names
(pkgsrc/devel/libsigc++2 is a good example), do a tar -tf to get
a list of filenames, compress it and do a tar -tzf to get another
list of the filenames. Notice that the two lists differ.

Propery handle "cpio" archives where the last hardlink includes the
data of a file. This fixes PR bin/26514.

Salvage the `create parent directories and try again' path which is
removed by the previous commit.

break out of the loop and avoid the infinite loop if attempts to
create, unlink and create all fail.

don't attempt to remove the current directory.

Completely rework how tools/compat is done. Purge all uses/references to
_NETBSD_SOURCE as this makes cross building from older/newer versions of
NetBSD harder, not easier (and also makes the resulting tools 'different')

Wrap all required code with the inclusion of nbtool_config.h, attempt to
only use POSIX code in all places (or when reasonable test w. configure and
provide definitions: ala u_int, etc).

Reviewed by lukem. Tested on FreeBSD 4.9, Redhat Linux ES3, NetBSD 1.6.2 x86
NetBSD current (x86 and amd64) and Solaris 9.

Fixes PR's: PR#17762 PR#25944

Don't exit with 1 on broken archives that mark directories as files for
the benefit of pkgsrc. I would have prefered that people fixed their
archivers instead.

Set the chflags *after* the rename, not before.

PR/25350: Alan Barrett: Treat the sticky bit specially as non-superusers
cannot set it on non-directories (elvis-2.2_0 extraction failed)

PR/25188: Thomas Klausner: pax-as-tar problem with trailing slashes
The problem is with the program that generates the tar file:
-rwxrwxrwx 1 root wheel 0 Feb 8 16:46 faad2/aacDECdrop/
It creates directory nodes without the 'd' bit set, so that pax thinks
they are files and does the temporary name and dance with them. Added
code to detect this condition, warn about it, and work around it.

PR/25207: Hubert Feyrer: tar cores on certain archive contents
Don't forget to cleanup xtmp_name when we fail to create the temporary file.

branches: 1.41.2;
When pax catches a signal and cleans up, make sure to remove any temporary
file that was being extracted to. This will prevent pax from leaving
droppings when you hit ^C.

Nuke extra mkstemp. Use name, not tmp_name with chk_path.

Back out last change. Since the name is the entire path, prefixing it
with a '.' will not have the desired effect.

When creating the temporary file, prefix the name with a '.'

When restoring files from an archive, initially restore to a temp-file and
once it has been instantiated correctly, rename it to desired name. This
prevents the problem of partially created files being accessed before they
are complete. If said file is a shared library, that can cause ramdon core
dumps.

Overhaul how `build.sh tools' are used:

* Rename "config.h" to "nbtool_config.h" and
HAVE_CONFIG_H to HAVE_NBTOOL_CONFIG_H.
This makes in more obvious in the source when we're using
tools/compat/config.h versus "standard autoconf" config.h

* Consistently move the inclusion of nbtool_config.h to before
<sys/cdefs.h> so that the former can provide __RCSID() (et al),
and there's no need to protect those macros any more.

These changes should make it easier to "tool-ify" a program by adding:
#if HAVE_NBTOOL_CONFIG_H
#include "nbtool_config.h"
#endif
to the top of the source files (for the general case).

Move Keith Muller's code from a 4-clause to a 3-clause licence by removing
the advertising clause. Diffs provided in PR 22397 by Joel Baker, confirmed
to the board by Keith Muller.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22249, verified by myself.

no need for HAVE_LCHOWN and HAVE_LCHMOD anymore.

add missing braces.

from HIRATSUKA Kouichirou in PR bin/20276.

cosmetic error condition improvements.
- sprinkle some fflush() to print filenames correctly.
- print a \n to stdout to fix formatting.
- English, consistently use "Cannot" in error messages.

some whitespace cleanup.

Fix various spelling, grammar, and punctuation problems in comments, from Sergey Svishchev in PR 20164.

The Double-Semi-Colon Police.

writable, not writeable.

Fix oops in last, chown() failure should be fatal.

Noted by Charles Hannum.

Don't attempt to chown files if euid != 0, and make chown() non-fatal
if it does fail.

This addresses my PR bin/18912 (running pkg_info as non-root user on a
binary package).

Add proper longlink support. Previously we handled longname support, and
the longlink support was completely wrong.

merge OpenBSD changes:
- correct -C processing
- add ability to read filenames and flags from a file
- don't print dangerous escape sequences to the terminal
- use strlcpy/strncpy properly.
- handle tmpfile creation better.
- improve documentation of options.
- handle stdout/stderr list selection correctly.
- kill gzip when we get interrupted.
- simplify gzip setup.
- add more flags to programs.

additional changes:
- librmt processing.
- set POSIXLY_CORRECT in options parsing.
- prevent more string overruns.
- support -T

we don't turn the switch on to replace tar and cpio yet.

branches: 1.23.2;
"doesn't" should have an 's'.

Protect __RCSID and __COPYRIGHT from being invoked if not defined.

Make almost all tools compile and run properly on non-NetBSD hosts. (In
particular, most tools now run correctly on Solaris 7.)

- ansi KNF (just remove all the #ifndef __STDC__ prototype junk,
leaving the ansi stuff)
- use longlong_t instead of quad_t (etc), and rename *uqd*() -> *ull*()
- clean up the NET2_STAT stuff similar to ftpd; provide #defines and
macros which select which cast to use, etc
- clean up the NET2_FTS and NET2_REGEX #define use

minor KNF nits.

branches: 1.18.2;
Fix doubled 'the' in comment.

Cleanup and fix typos.
Partially from FreeBSD.

Fix SIGINFO botch (PR #8868).
Continue partial write(2) on signals (xwrite()).
Partial read(2) at a few places are also continued (xread()).

Add {read,write}_with_restart() hooks for porting on systems
which don't restart interrupted read()/write() calls.

Reviewed and discussed in tech-kern and tech-userlevel lists.

branches: 1.15.2;
Completely disable all the chflags(2) support for now. With this enabled, I
can't even get a build to finish over NFS. Not acceptable.

Don't try to set file flags for a symlink.

add a new -p subflag: 'f' to preserve 4.4BSD fileflags. this only
really works for `pax -rw' but that's where it is most useful.

support ././@LongLink extraction, as created by GNU tar.

remove some old cruft.

branches: 1.10.4;
fix typo in a comment

Deal with GCC warnings.

When copying time stamps, copy the microseconds, too. Also nuke silly
static structure.

Restore owner/group/mode/atime/mtime of symbolic links.

- Rename local warn so tty_warn that we can include <err.h> since this program
uses err() and errx().
- Fix printf format strings.
- Added WARNS=1; note
- Added missing prototypes.
- Split overloaded trail function into trail() and subtrail().

kill 'register'

convert to new RCS id conventions.

Fix up RCS ids.

Add RCS ID's

branches: 1.1.1;
Initial revision

From 4.4 lite

Pull up to last week's -current.

Completely disable all the chflags(2) support for now. With this enabled, I
can't even get a build to finish over NFS. Not acceptable.

file file_subs.c was added on branch comdex-fall-1999 on 1999-11-07 15:48:25 +0000

Sync w/ netbsd-1-5-base.

Pull up revisions 1.42-1.46,1.48-1.50 (requested by cube in ticket #1736):
Fix handling of certain archives, fixing core dump (PR#25207).
Fix problem with trailing slashes (PR#15188).
Treat sticky bit specially for non-superuser (PR#25350).
Set chflags after rename, not before.
Don't exit with 1 on broken archives that mark directories as files.
Don't attempt to remove current directory.
Break out of loop and avoid infinite loop in certain error situations.
Salvage the `create parent directories and try again' patch.

fix botched pullup from ticket #1021 and make sure this compiles on
non-NetBSD hosts again. Also fixes PR#24481

Pullup rev 1.24-1.41 (requested by rafal in ticket #1021)

Pullup pax to current version on trunk. Includes many fixes.

branches: 1.41.2.8.2;
Pullup rev 1.52 (requested by christos in ticket #921)

Keep track of the length of the gnu long name/link hack, so that
we can append to it. Modify the code so that file_write is re-entrant,
even in the gnu long name/link hack.

Pullup rev 1.51 (requested by tron in ticket #784)

Propery handle "cpio" archives where the last hardlink includes the
data of a file. PR#26514

Pullup rev 1.48-1.50 (requested by cube in ticket #754)

Fixes for pax breaking on some archives

Pull up revision 1.47 (requested by jmc in ticket #527):
Completely rework how tools/compat is done. Purge all uses/references to
_NETBSD_SOURCE as this makes cross building from older/newer versions of
NetBSD harder, not easier (and also makes the resulting tools 'different')
Wrap all required code with the inclusion of nbtool_config.h, attempt to
only use POSIX code in all places (or when reasonable test w. configure and
provide definitions: ala u_int, etc).
Reviewed by lukem. Tested on FreeBSD 4.9, Redhat Linux ES3, NetBSD 1.6.2 x86
NetBSD current (x86 and amd64) and Solaris 9.
Fixes PR's: PR#17762 PR#25944

Pullup rev 1.45-1.46 (requested by snj in ticket #252)

Set the chflags *after* the rename, not before.
Don't exit with 1 on broken archives that mark directories as files for
the benefit of pkgsrc.

Pull up revision 1.44 (requested by snj in ticket #217):

PR/25350: Alan Barrett: Treat the sticky bit specially as non-superusers
cannot set it on non-directories (elvis-2.2_0 extraction failed)

Pull up revision 1.43 (requested by snj in ticket #217):

PR/25188: Thomas Klausner: pax-as-tar problem with trailing slashes
The problem is with the program that generates the tar file:
-rwxrwxrwx 1 root wheel 0 Feb 8 16:46 faad2/aacDECdrop/
It creates directory nodes without the 'd' bit set, so that pax thinks
they are files and does the temporary name and dance with them. Added
code to detect this condition, warn about it, and work around it.

Pullup rev 1.42 (requested by christos in ticket #170)

Fixes for multiple pax pr's. PR#25200. PR#25207.

Apply patch (requested by riz in ticket #5512):
Sync pax with HEAD of 2005-07-22.

sync with HEAD

Sync with HEAD.

Third (and last) commit. See http://mail-index.netbsd.org/source-changes/2009/05/13/msg221222.html

Pull up following revision(s) (requested by christo in ticket #678):
bin/pax/extern.h: revision 1.56
bin/pax/file_subs.c: revision 1.62
bin/pax/options.c: revision 1.104
bin/pax/pat_rep.c: revision 1.29
bin/pax/pax.h: revision 1.30
Fixes from christos for pr-41167
XXX -X is still broken, but that wasn't part of this PR.

Rebase to HEAD as of a few days ago.

sync with head.

for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.

this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")