108 lines
5.1 KiB
Plaintext
108 lines
5.1 KiB
Plaintext
SHORTTERM
|
|
- handle international filenames more gracefully (unicode API?)
|
|
- most is multithreaded ... but zzip_dir_open (Thorsten Schöning)
|
|
- rboerdijk@ does also report errors on overlapping reads, another
|
|
one pointed to the usage of seek_set that may cause the problems
|
|
|
|
WISHLIST
|
|
|
|
- Check the CRC value at the end of read... and add more error codes.
|
|
|
|
- Do more test. Currently use only with tested "friendly" archives.
|
|
This is also related to usages of zziplib in virus detection
|
|
code which should better have a hardened library code. That does
|
|
also include inflate interface code to need the most testing.
|
|
|
|
- the buffer reusage code was not strictly multithreading. It should
|
|
be fixed by now but it would be better to have an automatic test
|
|
routine to check reentrancy/multithreaded functionality.
|
|
|
|
- Sligthly More documentation. With the generation of man pages and
|
|
multiple pages for the website, it does already look acceptable.
|
|
It should still get better of course - kinda newbie friendly *g*
|
|
|
|
- Boris Schäling likes to open a zzip archive in memory.
|
|
|
|
KNOWN PROBLEMS
|
|
|
|
The win32 compilers need each a different config.h derivate that
|
|
matches both the headers shipped with the compiler and installed
|
|
with updates of the SDK. There is no autoconfigure on win32 as
|
|
that - unless you install some unix tools along.
|
|
|
|
The sparc-sun-solaris2.* will utter warnings for "char subscript"
|
|
which is caused by isdigit() from ctype.h - this will NOT FIX as
|
|
it is only in the example source code and we want to keep those
|
|
lean and mean to make them easy to adopt by developers.
|
|
|
|
The hppa1.1-hp-hpux10.20 did show spurious problems of making
|
|
shared libraries - this may well fix with an update of the
|
|
libtool package, the libtool 1.4 is dated 2001/04/24
|
|
|
|
There are reports of misaligned access to some zip fields that
|
|
I would guess to be on little-endian non-x86 platforms. The current
|
|
bytewise access of multibyte fields is targetted towards the
|
|
bigendian unix machines. The fix would need to go to fetch.h but
|
|
so far no response came about as that one could test a solution.
|
|
|
|
There are spurious reports of users on win32 platforms that tell
|
|
of some problems with a specific zip file they have but it was
|
|
not possible so far to recreate an environment abroad to show
|
|
the problem too. One can not say if that is due some general
|
|
instability out of DLL hell, or if there is a bug hiding somewhere.
|
|
Please send all those zip files to the maintainer, perhaps it
|
|
can help to find the real cause (I doubt it is in zziplib, but..)
|
|
|
|
Since lately the xml docbook tools have hardened the checks on the
|
|
input xml that is used for manpage generation. Interestingly the
|
|
resulting manpages are still okay but one should try to fixaway the
|
|
warnings as may be later the result would lead to garbage output
|
|
due more changes in the tools. Needs to change the xml generator
|
|
used in zzip (a python script).
|
|
|
|
TESTED PLATFORMS
|
|
sparc-sun-solaris2.6/gcc2.95.3
|
|
sparc-sun-solaris2.8/gcc2.95.3
|
|
hppa1.1-hp-hpux10.20
|
|
i686-mandrake-linux-9.0/gcc3.2
|
|
i686-mandrake-linux-9.1/gcc3.2.2
|
|
i686-debian-linux-2.2/gcc2.95.2
|
|
i386-unknown-freebsd4.7/gcc2.95.4 (formerly with wrapwrap)
|
|
powerpc-apple-darwin5.5 (formerly with wrapwrap)
|
|
alphaev67-unknown-linux-gnu/gcc2.95.4 (that's a 64bit platform)
|
|
i386-ms-win32/msvc6
|
|
i386-ms-win32/msvc7
|
|
i386-ms-win32/mingw+msys
|
|
... and probably a lot of others not known to the maintainer.
|
|
|
|
Additionally, note that Sourceforge has discontinued their compilefarm
|
|
server laboratory. That makes it unlikely that proper support for
|
|
crossplatform functionality can be provided. Expect a compile problem
|
|
here or there - the code however should be prepared to get around any
|
|
problems easily. Send patches! (especially Linux distributions makers
|
|
are usually not sending their patches to upstream maintainers).
|
|
|
|
Note: the latest cross platform tests are done indirectly by using
|
|
the build.opensuse.org rpm packaging where one can run "make check"
|
|
just before doing the "make install" of the compiled library.
|
|
|
|
|
|
SUSE BUILDSERVER INFO
|
|
|
|
I: A function overflows or underflows an array access. This could be a real error,
|
|
but occasionaly this condition is also misdetected due to loop unrolling or strange pointer
|
|
handling. So this is warning only, please review.
|
|
W: zziplib arraysubscript ../../zzip/memdisk.c:114
|
|
|
|
I: File is compiled without RPM_OPT_FLAGS
|
|
W: zziplib no-rpm-opt-flags <cmdline>:../../SDL/SDL_rwops_zzcat.c, ../../SDL/SDL_rwops_zzip.c
|
|
|
|
I: Program is likely to break with new gcc. Try -fno-strict-aliasing.
|
|
W: zziplib strict-aliasing-punning ../../zzip/file.c:275
|
|
W: zziplib strict-aliasing-punning ../../zzip/fseeko.c:99, 126, 147, 158, 182, 281, 288, 301, 360, 539, 543, 546, 563
|
|
W: zziplib strict-aliasing-punning ../../zzip/memdisk.c:181, 182, 183, 185, 186, 187, 188, 189, 192, 193, 194, 195, 247, 455, 456
|
|
W: zziplib strict-aliasing-punning ../../zzip/mmapped.c:277, 289, 311, 314, 339, 340, 393, 397, 410, 438, 440, 443, 444, 549, 551, 552, 558, 559, 561
|
|
W: zziplib strict-aliasing-punning ../../zzip/zip.c:318, 320, 321, 322, 339, 341, 342, 343, 484, 485, 486, 497, 498, 499, 500, 501
|
|
E: zziplib 64bit-portability-issue ../../zzip/memdisk.c:112
|
|
System halted.
|