Tim Kientzle [Tue, 11 Aug 2009 15:55:52 +0000 (11:55 -0400)]
Probe acl_get_link and acl_get_link_np so that ACLs are properly
read from symlinks. Without this, link-related tests fail for cpio
on FreeBSD. (Same as r1357, but for autoconf-generated builds.)
Tim Kientzle [Tue, 11 Aug 2009 15:14:32 +0000 (11:14 -0400)]
Fix cpio symlink tests on FreeBSD; the link-sensitive ACL routines
weren't being properly configured, which was causing the regular
Posix mode bits to get overwritten.
Tim Kientzle [Fri, 7 Aug 2009 04:01:06 +0000 (00:01 -0400)]
Test whether printf() supports %jd, %ju, %lld, or %llu.
This needs to be implemented in the autoconf machinery
and tar and cpio need to be taught to use it when printing
very large numbers.
Tim Kientzle [Thu, 6 Aug 2009 05:29:00 +0000 (01:29 -0400)]
Fix bsdcpio build on VS9. There's still a minor build issue on MinGW that I need to track down. I think some of the cpio_windows.c shims aren't needed and can be removed.
Tim Kientzle [Sun, 2 Aug 2009 21:41:53 +0000 (17:41 -0400)]
Reduce diffs between tar/test infrastructure and cpio/test infrastructure.
In particular, this makes it possible to use the new platform-independent
assertions for creating dirs, testing file attributes, etc.
Tim Kientzle [Sun, 2 Aug 2009 20:00:46 +0000 (16:00 -0400)]
Rework the style of CHECK_FUNCS() to better match
the style of built-in cmake tests:
* One test per invocation
* macro name is explicit
Convert some of the tests to CHECK_SYMBOL_EXISTS(),
which further reduces some of the magic within the
old CHECK_FUNCS. It is now primarily an extended
version of CHECK_FUNCTION_EXISTS that understands the
glibc __stub_ convention for marking functions that
should be treated as nonexistent even though they
do exist in the local library. I've renamed the
macro to CHECK_FUNCTION_EXISTS_GLIBC to better
reflect this.
Tim Kientzle [Sun, 2 Aug 2009 19:16:28 +0000 (15:16 -0400)]
Revert r1289 and correct some of the *id_t defines:
* uid_t/gid_t/id_t all default to "short" on Win32, "unsigned int" elsewhere
* pid_t doesn't need to be probed so remove it
Tim Kientzle [Mon, 27 Jul 2009 04:27:03 +0000 (00:27 -0400)]
Trim down Windows shims for tar: A lot of this has simply gone away:
* the open/read/write/close emulations were for the test harness, which now uses stdio,
* mkdir() was for the test harness, which has been refactored so it's no longer needed here
* opendir/readdir/closedir are only needed by tree.c and are being pushed down into tree.c
Tim Kientzle [Sun, 26 Jul 2009 04:12:29 +0000 (00:12 -0400)]
Merge various test framework improvements across the various
test harnesses. In particular, there have been a few spot
changes to improve Windows support.
Tim Kientzle [Sun, 26 Jul 2009 02:59:29 +0000 (22:59 -0400)]
Provide assertMakeDir() for tests to use instead of raw mkdir(). In particular, this allows us to hide the difference between POSIX two-argument mkdir() and Windows single-argument version.
Charles Wilson [Tue, 21 Jul 2009 04:54:01 +0000 (00:54 -0400)]
Minor MinGW fixes.
* bsdtar_windows.h: ensure struct bsdtar and struct archive are forward
declared before declaring functions that take pointers to those types.
Charles Wilson [Tue, 21 Jul 2009 04:52:25 +0000 (00:52 -0400)]
Minor MinGW fixes.
* bsdtar_windows.h: New MinGW already defines PRId64; guard it.
Ensure we have definition of MAX_PATH in time, by including windows.h.
Charles Wilson [Tue, 21 Jul 2009 04:43:29 +0000 (00:43 -0400)]
[MinGW/MSVC] Use correct types for (replacement) uid_t and git_d
The st_uid and st_gid fields in struct stat are shorts, not unsigned
int nor (regular) int for MinGW (and, AFAICT, for MSVC as well).
However, neither platform defines uid_t or gid_t, but AC_TYPE_UID_T
defaults to "int" in that case, and the current defaults for __LA_UID_T
and __LA_GID_T in archive.h and archive_entry.h are "unsigned int". On
MinGW these ('int' and 'unsigned int') do not match each other, and they
certainly do not match 'short'. Fix it.
* build/autoconf/la_uid_t.m4: New file defines la_TYPE_UID_T replacement
macro for AC_TYPE_UID_T.
* configure.ac: Use it.
* libarchive/archive.h [_WIN32]: Use correct default values for
__LA_GID_T and __LA_GID_T.
* libarchive/archive_entry.h [_WIN32]: Ditto.
* libarchive/archive_write_set_format_pax.c (archive_write_pax_header):
Avoid size-of-comparison warnings when uid_t/gid_t have less than 18 bits.
Tim Kientzle [Sun, 19 Jul 2009 07:57:34 +0000 (03:57 -0400)]
Sort short option list, fill in missing -I option, update
manpage options to include -0, -J, -j, -lzma, and -h.
Since Windows no longer tries to automatically turn on ownership
restore, we need --preserve-owner to allow users to manually
turn it on.
Tim Kientzle [Sun, 19 Jul 2009 07:20:51 +0000 (03:20 -0400)]
Remove the attempt to determine "root" for Windows.
It's not really feasible and just gets weirder the longer
we look at it. We now assume different defaults only
for the root user on non-Windows systems.
Tim Kientzle [Sat, 18 Jul 2009 05:13:51 +0000 (01:13 -0400)]
More simplification of the seek/truncate logic:
1) We don't actually need to truncate; seek is enough.
2) The Windows code doesn't need to be more paranoid than the Posix code.
projects / thirdparty / libarchive.git / log
