MFp4: libarchive 2.5.4b. (Still 'b' until I get a bit more

feedback, but the 2.5 branch is shaping up nicely.)

In addition to many small bug fixes and code improvements:
 * Another iteration of versioning; I think I've got it right now.
 * Portability:  A lot of progress on Windows support (though I'm
   not committing all of the Windows support files to FreeBSD CVS)
 * Explicit tracking of MBS, WCS, and UTF-8 versions of strings
   in archive_entry; the archive_entry routines now correctly return
   NULL only when something is unset, setting NULL properly clears
   string values.  Most charset conversions have been pushed down to
   archive_string.
 * Better handling of charset conversion failure when writing or
   reading UTF-8 headers in pax archives
 * archive_entry_linkify() provides multiple strategies for
   hardlink matching to suit different format expectations
 * More accurate bzip2 format detection
 * Joerg Sonnenberger's extensive improvements to mtree support
 * Rough support for self-extracting ZIP archives.  Not an ideal
   approach, but it works for the archives I've tried.
 * New "sparsify" option in archive_write_disk converts blocks of nulls
   into seeks.
 * Better default behavior for the test harness; it now reports
   all failures by default instead of coredumping at the first one.

1 files changed, 5 insertions, 0 deletions

diff --git a/lib/libarchive/test/test_tar_large.c b/lib/libarchive/test/test_tar_large.c
index 73113cd..b0a277f 100644
--- a/lib/libarchive/test/test_tar_large.c
+++ b/lib/libarchive/test/test_tar_large.c
@@ -242,6 +242,11 @@ DEFINE_TEST(test_tar_large)
 		archive_entry_copy_pathname(ae, namebuff);
 		archive_entry_set_mode(ae, S_IFREG | 0755);
 		filesize = tests[i];
+
+		if (filesize < 0) {
+			skipping("32-bit off_t doesn't permit testing of very large files.");
+			return;
+		}
 		archive_entry_set_size(ae, filesize);
 
 		assertA(0 == archive_write_header(a, ae));