On macOS, shared libraries end in "$(SOVERSION).dylib", not
".so.$(SOVERSION)" like on Linux. Also, a different linker option is
needed to set the equivalent of the soname.
This is based on patches from:
Elmar Pruesse <elmar.pruesse@ucdenver.edu>
and
Roy Storey <kiwiroy@users.noreply.github.com>
but refactored, and fixed to not break the Windows build.
Apparently, mmap'ing /dev/zero fails on macOS. So contrary to the
intent, it's not actually portable. Switch to MAP_ANONYMOUS instead,
with a fallback to MAP_ANON if it's not defined.
The API returns the compressed size as a size_t, so
deflate_flush_output() needs to return size_t as well, not u32.
Otherwise sizes >= 4 GiB are truncated.
This bug has been there since the beginning.
(This only matters if you compress a buffer that large in a single go,
which obviously is not a good idea, but no matter -- it's a bug.)
Resolves https://github.com/ebiggers/libdeflate/issues/44
libdeflate's 'gunzip' program fails if the uncompressed file size is a
nonzero multiple of 4 GiB because then the ISIZE field in the gzip file
is 0, so gunzip allocates no space for the decompressed data. Then when
libdeflate returns LIBDEFLATE_INSUFFICIENT_SPACE, gunzip doubles the
buffer size which stays 0, causing it to report "file corrupt or too
large to be processed by this program".
While neither libdeflate nor its 'gunzip' program is designed for single
gzip streams this large anyway, this particular bug can easily be fixed
by always allocating at least one byte for the decompressed data.
Add the user-specified CFLAGS after the default optimization flags
rather than before, so that the default optimization flags can be
overridden.
[EB - updated comment and improved commit message]
The Makefile forced installation into $(PREFIX)/lib, e.g. /usr/lib.
But many distros use /usr/lib64 instead.
Therefore, allow specifying a custom BINDIR, INCDIR, and/or LIBDIR.
This allows the parent shell to call on Gentoo Linux:
make install DESTDIR="${DESTDIR}/${EPREFIX}" LIBDIR=/usr/"$(get_libdir)"
[EB - improved commit message]
Move program utility functions that are used only by "test programs"
(i.e. not by gzip/gunzip) from prog_util.{c,h} into test_util.{c,h}.
This reduces the code that is compiled for the default build target,
which excludes the test programs.
Another build_decode_table() optimization: rather than filling all the
entries for each codeword using strided stores, just fill one initially
and fill the rest by memcpy()s as the table is incrementally expanded.
Also make some other cleanups and small optimizations.
Further improve build_decode_table() performance by splitting the "fill
direct entries" and "fill subtable pointers and subtables" steps into
separate loops and making some other optimizations.
Improve libdeflate's worst-case performance decompressing malicious
DEFLATE streams by about 14x, bringing it within a factor of about 2x of
zlib, by skipping rebuilding the decode tables for the static Huffman
codes when they're already loaded into the decompressor.
This improves performance decompressing a stream of all empty static
Huffman blocks from about 0.36 MB/s to 175 MB/s, or the original
reproducer given on the Github issue from about 3.3 MB/s to 219 MB/s.
A regression test is added for these cases as well as the empty dynamic
Huffman blocks case to verify worst-case performance comparable to zlib.
Resolves https://github.com/ebiggers/libdeflate/issues/33
