Now hosted on github.com/gdraheim/zziplib
2018-02-02  guidod  <guidod@gmx.de>
    * fix a number of CVEs reported with special *.zip files
    * the testsuite has been expanded to cover all the CVEs
    * some minor doc updates referencing GitHub instead of sf.net

2017-06-11  guidod  <guidod@gmx.de>
    * fix portability issues by introducing <zzip/__errno.h>
    * fix portability issues by introducing <zzip/__mkdir.h>
    * fix portability issues by introducing <zzip/__string.h> for strndup
    * reuse portability header <zzip/__fnmatch.h> in bins/*.c
    * release v0.13.67

2017-04-24  guidod  <guidod@gmx.de>
    * common frontend 'unzzip.c' for example code in unzzip*-xx.c
    * expand testcases to cover all of big/mem/mix/zap variants
    * release v0.13.66

2017-04-23  guidod  <guidod@gmx.de>
    * turn the testsuite from makefile into python unittest format
    * extend 'make check' testsuite to cover non-libzzip extensions
    * fix unzzipcat-seeko and unzzipcat-mem to match the testsuite
    * release v0.13.65

2017-04-20  guidod  <guidod@gmx.de>
    * check patches in opensuse builds, mostly for some CVE reports

2012-09-15  guidod  <guidod@gmx.de>
    * zzip.h, plugin.h: "};" at end of extern-C produces build errors with
       the default --pedantic-errors on newer CMake. Thanks to Keith Thompson
       recognizing it - see http://stackoverflow.com/questions/12384280/ 

2012-03-11  guidod  <guidod@gmx.de>

	* configure.ac: fallback to libtool -export-dynamic unless being sure to
	   use gnu-ld --export-dynamic. The darwin case is a bit special here 
	   as the c-compiler and linker might be from different worlds.
	* next 0.13.62 - needs a fix for mingw 
	   (-mconsole is not a linker option, only the gcc should see it)

2010-12-30 guidod <guidod@gmx.de>
        * done 0.13.60 release, tagged, and announced.
        * prepare 0.13.61

2010-12-29 guidod <guidod@gmx.de>
        * Makefile.am: allow nonstaic build
        * wrap fd.open like in the Fedora patch

2010-12-28 guidod <guidod@gmx.de>
        * zziplib.spec - simplify the spec based on distro specs.

2010-12-27 guidod <guidod@gmx.de>
        * adding fetch.h/fetch.c to msvc8/zziplib.vcproj to allow
          tests with Visual Studio C++ Express 2005

2010-12-27 guidod <guidod@gmx.de>
        * fix stat.c to check ZZIP_CASELESS instead of the deprecated
          ZZIP_CASEINSENSITIVE. The latter is still recognized for
          compatibility reasons and implicitly mapped to ZZIP_CASELESS.
        * fixes the zziplib-Patches-3078188 which reported the bug.

2010-12-27 guidod <guidod@gmx.de>
        * apply patches by Peter Breitenlohner for better compatibility
          with AIX host system. Also the Sun Studio compiler insists on
          static inline functions or else assumes extern inline.
        * also fixes zziplib-Patches-3011551 on the AIX problem.

2010-12-27 guidod <guidod@gmx.de>
        * add hints that the ext_io plugin handlers should be static.

2010-05-09  guidod  <guidod@gmx.de>

	* zzip/__fnmatch.h: add include <string.h> to make mingw32 happy.
	     Patches item #2995373 by Gregory Smith 

2010-05-08  guidod  <guidod@gmx.de>

	* SDL/Makefile.am: fix generation of sdl-uninstalled.pc that was
	   not recognized during compilation.

2010-04-05  guidod  <guidod@gmx.de>

	* Makefile.am: fix "osc rm" and fix "test-sdl" if SDL was not found.
	* configure.ac: SDL example builds on pkgconfig/sdl.pc which
	
	    was introduced with SDL 1.2.11 - CentOs5 ships 1.2.10 which
	    does only have "sdl-config" and it made for a mismatch in
	    in the build configuration: sdl-devel yes, pkgconfig sdl no.
	* zzip/zip.c (__zzip_parse_root_directory): move the check for
	     zzip_disk_entry_check_magic to be after the the check that
	     the seek-value is more than a header's size before the end.
	     (based on a segfault report by Tulipánt Gergely when the
	      new modulo_entries code is enabled (which is off by default))

2010-03-30  guidod  <guidod@gmx.de>

	* zzip/Makefile.am:  fixing symlinks for lib* - use only for Linux
	    Thanks to Ryan Schmidt (macports):
		... likes to install some symlinks that don't make sense on Mac OS X:
		$ ls -l /opt/local/lib/*zzip*so*
		lrwxr-xr-x  /opt/local/lib/libzzip*.so.10 -> libzzip*.so.13
		lrwxr-xr-x  /opt/local/lib/libzzip*.so.11 -> libzzip*.so.13
		lrwxr-xr-x  /opt/local/lib/libzzip*.so.12 -> libzzip*.so.13
		Yes, the filenames literally contain an asterisk.
	
    * prepare 0.13.60

2010-03-29  guidod  <guidod@gmx.de>

	* Makefile.am: change osc to new home:guidod:zziplib-13
	* configure.ac: update to autoconf 2.61 which does some things differently
	* m4/ax_pax_tar.m4: bug report http://trac.macports.org/ticket/24227
	  showed a problem on MacOsX where some "pax" tools where advertising
	  an "-O" option in their "--help" page but it did not work. Therefore
	  rewriting the macro to actually test some "-w -O" to create a tarfile.

2010-03-26  guidod  <guidod@gmx.de>

	* zzip/dir.c: and zzip/file.c - adding comments to the compat32
	   functions such that they appear near their off_t variants.
	   Also added the ifdef-EOVERFLOW check that is needed on Win32.

2010-03-25  guidod  <guidod@gmx.de>

	* Makefile.mk: add "-I m4" and re-autoconf with newest autotools.
	* zzip/zzip32.h: introducing compat32 functions that take "long"
	     arguments where the others use "off_t". Basically, there are
	     two calls per each off_t-function on all systems where 
	     previously that had been only done for those 64on32 systems. 
	     This scheme maximized binary compatibility - if an application
	     can not be compiled with _LARGEFILE_SOURCE then they can use
	     the compat32 variants to be always sure to do the right thing.

2010-02-15  guidod  <guidod@gmx.de>

	* zzip/Makefile.am: newest "install" does not like to install fetch.h
	                   three times ("...not overwite... it just installed")
	* Makefile.am: allow to build locally with "osc" (opensuse buildservice) 

2010-02-14  guidod  <guidod@gmx.de>

	* zzip/Makefile.am: use install-*-hook instead of install-*-local
	* docs/Makefile.am: PYTHONDONTWRITEBYTECODE=1 to ensure running
	   python with a read-only source code repository (dynamic workspaces)
	   NOTE: python -B was introduced with Python 2.6 (so use environment)
	* zzip/plugin.c: missing "void"
	   again thanks to Peter Breitenlohner (tex-live tug org)
	* test/Makefile.am: build-zips outside of source tree
	* m4/ax_prefix_config_h.m4: reduce error to warning as suggested
	* m4/ax_check_aligned_access_required.m4: merging with derived
	    macro zziplib-aligned-access.m4 by Peter Breitenlohner


2009-09-24  guidod  <guidod@gmx.de>
    * zziplib.spec: prepare 0.13.59
    * configure.ac: add --with-defines to allow compile-definitions
        for easy,harden,allow_modulo_entries (also for documentation)
	* zzip/zip.c: add ZZIP_EASY to disable the ZZIP_CORRECT_ROOTSEEK
	    feature
	* zzip/zip.c (__zzip_parse_root_directory): add parenthesis to fix
		ZZIP_ALLOW_MODULO_ENTRIES (as reported by Tulipánt Gergely)
	* zzip/fetch.h: add MSVC defines for x86/x64 that allows direct access

2009-08-23  guidod  <guidod@gmx.de>

	* Makefile.mk: remove some old stuff related to the compilefarm
	* zziplib.spec: prepare 0.13.58
    * add ports/symbian-S60/zziplib-symbian.zip (thanks to Rangel Reale)
	* switch from CVS to SVN (!!!!) and update some docs/*.htm files
	* zzip/mmapped.c (zzip_disk_mmap): compile error on win32-mingw [1771707]

2009-08-16  guidod  <guidod@gmx.de>

	* configure.ac: reorder ax_gcc_flags as the extra options were
	  disabled (do not know why) and add -Wstrict-aliasing to the set.
	  Then fix issues in source files (mostly signed-comparision warnings)
	* zzip/file.c (zzip_file_open): ZZIP_CASELESS will include a
	   mapping of "\" into "/" to get away with windows-orientied
	   programs. The backslash however is only active for those
	   windows' binaries. Fixes an issue with non-msvc win32-builds.

2009-08-04  guidod  <guidod@gmx.de>

        * m4/ac_sys_largefile_sensitive.m4: update, fixes ticket:2825391
	* zzip/mmapped.h: zzip/memdisk.h: replace buflen argument
	  from "int" to "size_t" type for init from some mem area.

2009-08-03  guidod  <guidod@gmx.de>

	* zziplib.spec: prepare 0.13.57
	* bins/zziptest.c: redefine to win32-Sleep() when seen WINDOWS_H
	  which allows it to compile also with a gcc/mingw/msvcrt target
	* bins/Makefile.am: ZZIP_LDFLAGS are not needed for *.exe targets
	  in fact they make for problems in crosscompiling without libtool
	* zzip/fseeko.c: typo "fseeko" -> "fseek" made for a compile error
	  on systems that do not have fseeko like windows msvcrt
	  (reported by Josh Heidenreich)

2009-07-03  guidod  <guidod@gmx.de>

	* zzip/mmapped.c: insert zzip_disk_buffer and
	* zzip/memdisk.c: zzip_mem_disk_buffer where both functions
	    will actually wrap an external buffer (e.g. mmapped from
	    another resource such as shared memory).
	* zzip/mmapped.c and zzip/mmapped.h: uncovered a bug for the
	   MATCH_NOCASE handling and WRAPPED_BUFFER handling that is
	   now handled by explicit ->flags values. The actual flag
	   values are now exported to the header instead of just using
	   a simple "1".
	* ideas for zzip_disk_buffer / zzip_mem_disk_buffer sprang from
	   a proposal by JP Dai [Goldleaf] who offerd very similar code.

2009-06-27  guidod  <guidod@gmx.de>

	* zziplib.spec: 0.13.56 - and make build.opensuse.org happy by
	   removing the %if/%endif sections. (it seems one buildserver
	   script understands it while the other don't - and effectivly
	   all variants fail if it is left in. Hopefully no other platforms
	   have problems with the additional buildrequires that are put here)
	* zzip/mmapped.c (zzip_disk_new): malloc(sizof()) was errornously
	    using a pointer instead of ZZIP_DISK. Not part of libzzip but
	    of the extra library libzzipmmapped that is in-memory-only.
	    (problem reported and solution shown by JP Dai [Goldleaf])
	* configure.ac: disable search for python2.2 (gentoo bug 275247+174189)

2009-06-22  guidod  <guidod@gmx.de>
    * zziplib.spec: 0.13.55 and integrate patches for build.opensuse.org
	* zzip/zip.c (__zzip_parse_root_directory): rework parse_directory
	   to allow for ZZIP_ALLOW_MODULO_ENTRIES version as required for
	   some cases where more than 65535 entries are packed into a
	   non-ZIP64 archive. The modulo_entries variant is disabled by
	   default but the know for the problem space was put into code to
	   check for two more error conditions and report them in debug mode.
	   (problem space was described carefully by Tulipánt Gergely and
	    the provided patch was a good foundation for the ifdef code)
	* configure.ac: use $can_build_shared from linker-config to allow
	    build on platforms that do not support a shared library concept
	    or where it is disabled (the commandline switch overrides it?)
	    (should fix the SF bug ID 2796485 reported by Giovanni Bechis)
	* zzip/file.c: errornous usage of currentfp before setting it to
	   the new fp value in zzip_seek(). In the good case it would seek
	   on the wrong file and in the bad case it goes off on sigbus/NPE.
	   The problem occurs only if two+ zips are accessed simultaneously.
	   (problem reported and patch kindly provided by Evgeniy Muhin)
	* docs/Makefile.am: fixup for newer xmlto may be to generate
	   a subdirectory "man3" inside of our output directory "man3"
	   and gard against usage of man3/man3 for the tarball unpacking.
	   (problem reported by Liu Qi [debian package maintainer])
	* zzip/__hints.h: add else-part for ZZIP_GNUC_PACKED
		(makes for problems with MSVC as reported by Denny Kunker)

2009-05-24 guidod <guidod@gmx.de>
    * zzip/fetch.c - remove SET/GET fetch definitions that already
       exist in zzip/fetch.h where they have been corrected lately.
    * docs/Makefile.am - change from install-sf via shell.sf.net to
       upload-sourceforge via web.sf.net - and alias to "www" shorthand
    * docs/referentials.htm - fix typo.
    * 0.13.54

2009-05-23  guidod  <guidod@gmx.de>
    * SDL/Makefile.am - remove dependency on zzip-config (and sdl-config)
       which do not exist any since quite a time - instead use pkg-config
       > Thanks to Michael Sterret for bringing it up
    * SDL/SDL_rwops_zzcat.c: the SDL_rwops example was fixed, as the
         nmemb/size of the sdl-read call were swapped over. The testsdl
         self check (based on zzcatsdl) works OK.
    * SDL/Makefile.am - the targe installation paths for the example
         was fixed to include/SDL_rwops_zzip - additionally there is now
         an SDL_rwops_zzip.pc pkgconfig script that will refer to the new
         location (it had been previously installed in include/SDL, ouch).
    * Note that the old zzip-sdl pkgconfig-files are gone now. Given that the
         example was not quite working, it is improbable that anyone had
         been using them in real code, so there's nothing breaking here.
    * create a new package SDL_rwops-devel in zziplib.spec that will catch
         the example code, SDL_rwops_zzip includes and SDL_rwops_zzip pkgconfig
    * append "make check" and "make test-sdl" to zziplib.spec for  minimal
         pre-installation check.
    * Makefile.am: fix zzip-postinstall
    * zziplib.spec: specifiy %version for the Provides:-clauses
    * configure.ac: do not let autoconfigured LDFLAGS/LIBS bleed into zziplib.pc
    * zip.c: ensure that the zzip64_trailer entries is being used and that the
        parse_root_directory function honours the full 64bit in the "entries" counter
        (rejecting the patch from Tulipánt Gergely which would allow to get away with
        non-standard zip archives that do use a normal zzip_trailer with a 16bit entries
        counter that happens to store only the modulo value of the real entries in the
        archive - zippers should fail beyond 65535 entries or switch to the ZIP64 format).
     * docs/referentials.htm - include reference to  http://www.ogre3d.org/
     * zzip/fseeko.c: check some more return values from fseeko/fread calls.
     * 0.13.53

2009-05-22  guidod  <guidod@gmx.de>
	* docs/Makefile.am: break a dual target into two lines as
	   suggested by ticket:2405440
	* m4/ax_cflags_no_writable_strings.m4 update with sed-call
	   suggested by ticket:2155649
	* m4/ax_check_aligned_access_required.m4 update for the cross_compiling
	   case with libpcap-list of host_cpu targets thare require aligned access
	   suggested by ticket:2479788
	* introduce ax_pax_tar.m4 to use *.tar format for manpages.tar (instead of .ar)
	* 0.13.52

2009-05-21  guidod  <guidod@gmx.de>

	* docs/Makefile.am: make install-docs install-man3 rules so that
	     a failing "ar x" (on OpenBSD/vax) will not kill the whole
	     configure / make / make install cycle. However it simply
	     means that there will be no manpages on such platforms.
	     May be OpenBSD/vax has a "gar" (gnu ar) installed somewhere?
	* closes Patches item #2716806, was opened at 2009-03-27 13:01
	* 0.13.51

2009-05-21 guidod <guidod@gmx.de>
    * zzip/fetch.h: ensure that either direct-bswap or direc-deref is only used
               on platforms that do not require aligned memory access. The
               latest report from Sylvain Beucler has shown an error a MIPS
               platform (named  http://www.freedink.org/ running on PSP).
    * zzip/fetch.h: replace _zzip_attribute(const) with defines from zzip/__hints.h
    * zzip/__hints.h: introduced ZZIP_GNUC_PACKED for __attribute__((packed))
    * zzip/format.h: replace _zzip_attribute(packed) with defines from zzip/__hints.h
    * zzip/conf.h: remove _zzip_attribute - not needed anymore.

2008-12-27 guidod <guidod@pc3>
    * ax_create_pkgconfig_info needs a definition for datarootdir (introduced
       in latest autoconf/automake making for an error in ogre3d configuration
       as reported by Ignaz Forster)
	* release 0.13.50

2008-12-24 guidod <guidod@pc3>
	* add the dir-zzip-* to the cvs repo (as used on some webs)
	* move "make rpm" to *.am (instead of *.mk)
	* amd64 should be in libdir=$prefix/lib64 (in the spec file)

2008-11-24 guidod <guidod@pc3>
	* BuildRequires: python (for %package doc)


2008-03-22  guidod  <guidod@pc3>

	* zzip/fetch.h: honor ZZIP_HAVE_ALIGNED_ACCESS_REQUIRED for the Linux bswap
	optimization. See also 443880@bugs.debian "SIGBUS on Sparc".

2008-03-03  guidod  <guidod@pc3>

	* TODO: update to recent problems. Need to check my mail stack next time.

2008-03-02  guidod  <guidod@pc3>

	* zzip/info.c: reverted.
	* also: note that the Makefiles have been update to the newest autotools.
	  The project itself has been imported to an Eclipse based CDT project.

2007-03-19
 * enforce indent by automatic indentic-check

2007-03-17
 * http://www.securitylab.ru/forum/read.php?FID=21&TID=40858&MID=326187
  zziplib vulnerability due to usage of strcpy
 * add indent-check for enforcing some source code style.

2007-02-01
 * remake manpages.ar / htmpages.ar - redistribute zziplib-manpages.ar
   instead while the rebuild mechanics become a lot simpler
 * unless --maintainer-mode, make mans / install-mans part of the default

2007-01-31
 * applying patches from Mike Frysinger
 * consider making the manpage install target part of the default
   `make install` rather than a sep install-man3 target ? (from Mike)
 * Michael Sterret reports, that the SDL/* example requires zzip-config
   instead of the pkgconfig stuff - actually, the whole part should be
   give a makeover as newer SDL does install its own sdl.pc which makes
   the current local generation of zzip-sdl-config.pc a real mess.

2006-09-27
 * adding docs/zip-php.htm integrating Chris Branch's mail into the docs
 * appending docs/changes.htm rule from ./ChangeLog

2006-09-26
 * adding docs/notes.htm - to register some old dicussions for later
	reference. Let's see how that fills up.

2006-09-21
 * last doc/*.py updates
 * last doc/mksite.* updates
 * last doc/body.htm updates - including new zzipmmapped.html docu
 * "make install-sf" will copy the doctree directly to sourceforge

2006-09-18
 * docs/makedocs.py docs/zzipdoc/*.py creates new docs/zziplib-man3.tar.gz
	plus zziplib.html zzipmmapped.html zzipfseeko.html
 * that will be the only to be used in the future but it is not complete

2006-08-18
 * updating msvc8 project files
 * adding zlib.dll to cvs
 * zzip/memdisk.c:183 overrides for off64_t will not work with off32_t !!!
 * headers for mktime should return time_t - do we have the header ???

2006-08-17
 * A report by Tuilipant Gergely had shown an "unaligned access"
   error on an alpha machine. So, I have added macro to test for
   the condition of aligned access but the alpha platoform seems
   to be okay. Further reports are required.
 * Mark README.SDL to be explicitly outdated (it's 16. Dec. 2002)

2008-08-14
 * Yvan Phelizot reports a problem in zzip/mmapped.c
 * Malcom MacLeod had spurious problems inside a C++ project,
	so let's update all heraders with explicit 'extern "C"' linkage.

2006-08-08
 * rephrase build system, automated tests, some doc parts

2006-04-28
 * bsd/mac needs sys/types.h for size_t

2005-12-11
 * there have been reports about multithreading problems
 * one source of problems: access to the dir->cache members.
	fixed by using an explicit semaphore variable, otherwise
	just allocate/deallocate the buffer / filehandle
 * second source of problems: the zip file is opened _and_ read
	with only one filehandle for multiple threads that
	share the same "dir" handle. Look for "->currentfp".
 * that is not fixed away - while access to the cache variables
	can continue in the case of a "locked" state that is not
	possible for open/reads. In the "locked" case the thread
	must be blocked but that is a system-specific call.
 * there are two ways to fix it - (a) push down the sysfile to
	each zzip_file by "dup(2)"licating the sysfile handle.
	However zzip looses a feature that was helping a lot on
	system with a low number of sysfile hands (e.g. dos and
	some embedded operating systems that I did work with).
	and (b) provide an indirect ->lock() call that can be
	filled by the caller application upon zzip invokation,
	perhaps add it to the io-plugin structure. Well (c) is
	it possible to support both styles? Dynamically?

2005-12-10
 * testing on sourceforge compilefarm - including "make check"
 * there was an error on bsd'ish systems (implicit 64on32)
	which was caused by archaic CORRECT_ROOTSEEK code
	being still present. Remove it?

2005-12-09
 * testing with Microsoft Visual Studio 2005 (msvc8) Express Edition
 * fixing some compilation problems related to zip64 support in win32

2005-12-08
 * cut acinclude.m4 into separated aclocal macros in m4/ subdirectory

2005-10-14
 * testing on sourceforge compilefarm, cleaning away any compiler warnings
	i386-debian2.2 i386-freebsd4.8
	amd64-fedora3 alpha-debian3.0
	sparc-solaris8 powerpc-macos10.2
 * note, the build system has some quirks but not yet renovated to the
	newest autoconf/automake stuff anyway - TBD before final release

2005-10-13
 * MSVC7 knows the following __declspec attributes:
	align(#), allocate("seg"), deprecated, dllexport, dllimport,
	naked{for asm}, noinline, noreturn, nothrow, novtable,
	property{MngC++}, selectany{common}, thread{tls}, uuid("name")
 * introduce zzip__new__ for a restrict'ed return item
 * introduce zzip_byte_t to help silence down compiler warnings
	about signedness mismatch in assigning data buffer variables.
	Especially the zlib data_in happens to be of unsigned char type.
 * note that zzip/format.h does now use zzip_byte_t for layout definition

2005-02-17
 * adding that _GNU_SOURCE on __linux__ to get the strndup prototype
 * cleaning a few warning messages on different platforms as well
	as a bug for 64bit platforms
 * updating docs/mksite.sh and adding docs/mksite.pl and going to
	let VERSION be pasted directly instead of a makefile snippet
 * adding docs/memdisk.htm documentation to the series which is
	currently quite shallow.
 * note: no tests for actual zip64 support are made, neither for a
	large central directory (more than 64K files) nor for large
	files (more than 2GB). This is all theoretic but I am quite
	confident that it will work.

2005-02-16
 * change zzip/zip disk_trailer implementation
 * we do not anymore pass a copy of the file block around
 * we do use a local helper structure now with off_t fields
 * the parse_root_directory is now using off_t for all computatations
 * make check succeeds - beware, this is a new implementation
 * RANT: glibc bastards make for additinal warning messages, the fseeko
    prototype is not exported by default, but as soon as we add some
    define to make it available, some other prototypes are disabled,
    including one of the most widespread: strndup(). The documentation
    even says we have to say GNU to get prototype - if you have ever heard
    of unix to diverge into something not quite but be alike, here it is,
    what compiles on most platforms gives a warning on linux even that it
    works fine afterwards because the libc does export all we need and
    autoconf does find it - it is just not prototyped and that's all.

2005-02-15
 * adding zzip/memdisk.* as cache variant for libzzipfseeko
 * add macro support for zip64 extensions
 * extend trailer routines with zip64 detection (and return 0)
 * add memdisk bins

2005-01-04
 * remember TODO multithreading tests and hardening the code.
 http://www.idefense.com/application/poi/display?id=153&type=vulnerabilities

2004-12-27
 * updating to latest autoconf/automake/libtool stuff in Suse92
 * required a rewrite of ax_enable_builddir.m4 in ac-archive
 * adding two macros to silence down bogus depracation warnings
	from the latest gcc 3.3.5

2004-11-27
 * harveyandsu:yahoo:com reported an api mismatch for SDL rwops
	usage and the implementation in the SDL_rwops_zzip example.
 *

2004-05-11
 * documenting zzip-cryptoid handling
 * update mksite.sh

2004-05-09
 * documenting zzipmmapped and zzipfseeko parser libraries

2004-05-08
 * remove bogus zzip_file_open_ext_io from zzip.h
 * change to use mksite.sh for documentation builds

2004-03-08
 * add link in docs/history.htm to the new appnote.txt whitepaper
	on zip file format specification.

2004-02-19
 * a test run on solaris did show that fetch.* needs to be linked
	to the new lib*.la stuff - to convert endianess-dependent
	values from zip to main memory
 * there seems to be an automake problem with hpux10 that I can not
	yet define on what grounds it is - to get away with it I am
	not defining automake 1.7 as the version to be used, plus
	autoconf 2.57 - these are pretty new.

2004-01-24
 * zzip/fseeko.c and bins/zzip.c had some issued with non-C99 compilers
	which do not like variable declarations in the middle of a block

2004-01-16
 * zzip_rewind again - the deflateReset does not reset the
	input buffer variables (oops) so we have to do it
	explicitly. Here we adopt to set crestlen to full
	csize and the current z-buffer fillstate avail_in
	to null to trigger a new read() on next zzip_read.
 * the zlib headers do not tell whether next_in/avail_in
	is used in deflateReset as is done in deflateInit.
	A question to zlib@gzip.org brought no answer (as
	is always the case) and after some thinking I did
	decide to reset avail_in *after* the zReset call.
 * based on this decision, I have made up a patch and I was
	sending it over to zlib@gzip.org. Same consequence
	again, the message is bluntly ignored (while there is
	traffic on zlib devel mailinglist). There seems to be
	a major problem about it, either a technical problem
	 (mail-adress disfunctional) or a social problem (no
	one cares or perhaps some minor racism).

2003-12-29
 * added libzzipfseeko.la with another simplified interface
 * rename the other one to libzzipmmapped.la
 * fix bugs for decompression on the latter as well
   it seems the one has to use inflateInit2 (..., -MAX_WBITS)
 * added bins/unzzipshow for an example binary
 * added "--help" and "--version" to all bins examples

2003-12-24
 * added libzzipmmap.la with a simplified interface to zip archives
 * added bins unzzipdir and unzzipcat to show usage of libzzipmmap.la
 * introduce zzip/types.h to allow easier inclusion to zzip/mmapped.*

2003-12-22
 * add AC_C_BIGENDIAN and
 * move out ZZIP_GET to zzip/fetch.h where we also specify access
	wrappers to items in the zzip/format.h structures, portably.
 * add byteswap.h ac_check and put a new file autoconf.h to carry
	compiletime overrides for the autoconf definitions.
 * start changing over to use zzip/fetch.h definitions everywhere
	thereby replacing original ZZIP_GET16/32 usages completely.
	This is localized to zzip/zip.c
 * rename struct zzip_root_dirent to struct zzip_disk_entry but
	retain a compatibility declaration.
 * introduce zzip_version_t and zzip_dostime_t along with an optional
	ZZIP_NEED_PACKED def that helps with weird compiler struct packing
 * struct zzip_file_trailer did not use the char[x] variant, it does now
 * place even more zzip/fetch.h declarations
 * place even more defs into zzip/format.h
 * noting CVS area on webpages

2003-12-21
 * Nigel Stewart hints on some MSVC 6 issues, thanks.

2003-12-20
 * implant *Reset instead of Deinit/Init in zzip_rewind - the latter
	is supposed to make for a memory leak (see problem report).

2003-12-10
 * add zzip_rewind fix by glenn = Glenn Maynard (www.stepmania.com)
 * fix dirsep_casecmp buglet noted by glenn
 * glenn did ask for zzip_fstat, and here it is.
 * but glenn's file_dup is not as easy and skipped here.
 * otherwise, glenn noted that zziplib is not fully ready for multithreading,
	the reason seems to be mostly in the reshared 32k buffer where the
	access is not synchronized completely :-( ... effectivly, in MT
	all ZZIP_FILE*s need to have a dedicated ZZIP_DIR*buffer.
 * update make-doc.py with newest version

2003-08-18
 * changing license to dual MPL / LGPL
 * bumping version code 12.83 = 13.23 which are actually the very
	 same inside the code
 * remove a few compatibility headers at install time.
 * add install-exec-hook to add ln -s aliases for sharedlibs
	they are being tried again in .spec but that does not
	matter as long as it is "|| true"
 * adapt all source to carry the MPL alternative license
	instead of the old cruft.
 * remove staticlinking.txt - it should be covered by copying.htm now
 * remove "compats" disthook as well as the complete subdirectory
      i.e. rm -r "zziplib/" at top directory (actually, I store it
	in a distant place for now.)
 * remove _htmpages.ar and _manpages.are from dist tarball
 * add toplevel COPYING.LIB
 * adapt README and TODO
 * remove ac_output(zziplib and bins/zzip-config)
 * remove bins/zzip-config.in
 * adapt m4 macro and call zzip-config to be archaic
 * remove maintainer-mode for "test.zip"
 * turn zzip_plugin_io_t into a union reference, the old struct is
	still of the same name but a member in there called "fd"
	the typedef is no a "union _zzip_plugin_io" instead.
 * the member "use_mmap" is renamed to "sys" - we add a #define
	in plugin.h to let some software compile with all versions
 * in zzip/zip.c, try now with making maplen and pagesize of type
	zzip_ssize_t - as it was newly introduce it might help and
	perhaps be a little conservative in execution. Before that,
	we did use size_t and it was provoking warnings like
	comparing/computing signed+unsigned pairs of values.
 * add "type" and "write" members to zzip_plugin_io, keep it
	backward compatible as much as possible.
 * add _zzip_write to zzip/conf.h
 * remove ZZIP_WRAPWRAP code in plugin.c
 * remove --with-wrapwrap
 * introduce a #define _zzip_plugin_io_handlers and
	typedefs zzip_plugin_io_handlers and zzip_plugin_io_handlers_t
 * oops toplevel PHONY does not work, an automake restriction
 * oops, "cat body.htm" should be "cat $(srcdir)/body.htm"
 * adding manpages.ar again, just to be sure

2003-08-14
 * remove _not_implemend macro definitions for shallow-write api
	from header zzip.h
 * introduce symbol _ZZIP_WRITE_SOURCE that all files should
	carry that want to have the declarations of the
	shallow-write api
 * use that symbol for zzip/write.c and bins/zzip.c
 * adding docs/64on32.htm

2003-08-14
 * for the shallow-write api to compile on win32, we need to
	test for direct.h which only exists on win'sh system
	or so we assume
 * add bins/unzzip
 * add bins/zzip
 * ooops, zzip_strerror_of should check dir != 0

2003-08-13
 * complete zip-write API - no actual implementation so far
 * tested shallow-write implementation on solaris+linux+win32
 * and now, call it zzip-12 always

2003-08-12
 * `make docs` does not always work, so we add `make docu` now
 *  update make-doc.py
 *  fix call for make-doc.py
 *  fix zzip_get16 bad refer that gets detected by make-doc.py


2003-08-07
 * implement zzip/write to a degree that it forwards calls to real
	files and directories in the system. No zip archives here,
	but the accession API has been setup so that applications
	can start to pick it up even for versions of zziplib that
	have been compiled without zip archive support.

2003-08-07
 * add doc lines in zzip/zzip.h
 * add write-defines including flag-marker ZZIP_NO_CREAT
 * add file write.c with empty definitions for write-support
 * add the make-doc.py script in its last version - and make its output
	the default master files for both html and docbook
 * remove all other doc-maker scripts but make-doc.pl and make-doc.py
	plus the make-dbk.pl of course
 * twist configure.ac to set PYTHON instead of PHP

2003-08-03
 * make compatibility-headers to issue a warning upon include

2003-07-31
 * fix a segfault with corrupted central directory - just some checks
	for field boundaries
 * add a hack to support the fixup-rootdir thing with some simple(r) code
	that abuses a few fields in the trailer-structure
 * cvs checkout savannah.gnu.org:config into ../savannah.config and,
 * add a rule `make configsub` in Makefile.mk to update the two files
	uses/config.sub and uses/config.guess to the latest versions.

2003-07-30
 * creating a 0.12.82 out of the source for 0.10.82 and make up a rpm
	spec which does alias the .so files for these generations

2003-07-29
 * the include of _LARGEFILE_SOURCE to do ZZIP_LARGEFILE_RENAME is
	actually a bad idea - the respective symbol does not provoke
	64bit off_t on newer platforms. Only _LARGE_FILES seems to
	be correct here for older aix platform to get them define
	off_t = 64bit where others are using _FILE_OFFSET_BITS=64 now.
 * after changes to README, remember to package updated test.zip !!
	otherwise `make check` will fail.
 * modify _msvc.sed to include "define ssize_t int"
 * use ZZIP_EXPORTS instead of ZZIP_DLL for DLL export, that's the
	automatic way of MSVC 6 saying that "zzip" is a DLL project.
 * create new msvc6/zzip.dsp and delete msvc6/zziplib_DLL.dsp
 * all examples shall import from zzip.dll by linking with zzip.lib
	whereas that other zziplib.lib is just the static library
 * do a looong and detailed information in README.MSVC6
 * create a rule to pack the *.dll,*.exe,*lib files into a zip

2003-07-28
 * doc bug in zzip-zip reported by j.scrott.frank
 * remind me: unzip from memory seems to be more of a task lately...
 * forgot to modify RELASEINFO in rpm for zzip64 variant
 * also do not just use -D_LARGEFILE_SOURCE, it does not have the
	desired effect on all systems. Instead invent a new name
	-D_ZZIP_LARGEFILE to make it happen magically.
 * invent a rule to create zziplib64.pc from zziplib.pc for the
	renamed libzzip.so -> libzzip64.so
 * invent a rule to create zziplib64.la from zziplib.la for the
	renamed libzzip.so -> libzzip64.so
 * do also rename default libzzip.so to libzzip32.so and then create
	a symlink to it under the old name libzzip.so
 * create zziplib32.pc and zziplib32.la files accordingly. Following
	an application may choose explicitly between 32 or 64bit off_t
	version. This allows us to claim now that the default libzzip.so
	library may be either of them and not anymore strict 32bit off_t
 * aaahm, note that all of these account for rules in the linux rpm spec!
	other system do not get this packaging support from vanilla tarfile
 * well, better make it lib01 instead of lib0.10 for the rpm name
 * already roll out visualc7.zip and msvc6 files in subdirectories,
	do not ship them as .zip files
 * already adapt msvc6 .dsp files but not yet tested
 * rename main "rpm" target into "rpm2" to get rid of the warning message
 * use (cd zzip64 && make install) instead of (... make install-ltLIBRARIES)
	since otherwise we do not get any .pc file installed (oops)
 * while modifying *.la into *64.la, take care to fixup reference to the
	"old archive" *.a as well making it *64.a - same for *32.la file.
 * some file.c reports there was not previous declaration of some of
	its functions, but _only_ in 64bit compile step.
	Reason: looks like a gcc 3.2.2 bug, no fix here.
 * convert a few "ln -s" into "ln -s -f" in "make install" parts
 * update README to point out staticlink model
	in a separated LICENSE section.

2003-07-28
 * add three extra gcc options: -Wpointer-arith, -Wsign-compare, and
	-Wmissing-declarations which will help about portability to
	platforms where that is enabled by default in -Wall
 * sign-compare fix in, zzobfuscated (zzip_size_t -> zzip_ssize_t),
 * make 5 functions in zzipwrap/wrap.c global=>to=>static
 * add pre-delcare headers for 2 functions in zzip/zip.c
	__zzip_find_disk_trailer(), __zzip_parse_root_directory()
 * rewording of "ends/tail" in find-disk-trailer which does also find
	a little bug (ends-1 or not) in there, It should not have shown any
	practical problem so far, also corrected by anotgher little bug
	in there (ends-tail > sizeof or ends-tail >= sizeof)
 * adding a parts of a patch from Martin Schitter to harden zziplib
	for corrupted zip files.
 * remove a few duplicates in Makefile.mk, still need to keep "rpm"
 * add declaration for zzip_dir_alloc_ext_io into zzip/lib.h

2003-07-27
 * redefine size_t maplen -> off_t maplen and for each of the calls to
	mmap()/munmap()/read()/getpagesize() we cast to size_t - this might
	get us rid of a few warnings about signed/unsigned comparisons.
 * introduce AC_TYPE_SIZE_T and use the prefixed variant _zzip_size_t
	within plugin.h for io->read() calls. That makes for a little
	bit of portability - it seems some platforms (dot-net f.e.)
	do not `typedef unsigned int size_t;`, perhaps because int is 16bit
	and a 32bit or even 64bit entity is needed.
 * now we do not need to AC_COMPILE_CHECK_SIZEOF(size_t) anymore, perhaps
	even stand io-wrap is not needed anymore. Let's see.
 * both changes need update of zzip/conf.h for _zzip_size_t plus
	- plugin.h for io->read(,,size_t)
	- zzip.h for zzip_file_read (,,size_t)
	- zzip.h for zzip_read(,,size_t)
	- zzip.h for zzip_fread(,size_t,size_t,)
	- zzip.h for zzip_seek(,off_t,)
	- zzip.h for zzip_tell(,off_t,)
	- file.h int => size_t for zzip_file's restlen crestlen usize csize
 * the io->seeks(,off_t,) has not been respected so far sometimes.
	and like lseek(2), we define the returntype of zzip_seek as off_t
 * introduce zzip_ssize_t as well, and modify
	- zzip.h for zzip_read and zzip_fread and zzip_file_read
	- plugin.h for io->read() return type
 * dot-net bug in aligned4() - sizeof(long) < sizeof(void*) !!
	better use off_t in the hope that it does fix it
 * tested on ia32-linux, alpha64-linux, sparc32-solaris
 * ooops, we need to update io-plugin examples and documentation as well,
	since we define a custom _read() routine righ there.

2003-07-26
 * updating to AX_SPEC_DEFAULTS macro
 * zzshowme does not work with no libzzip installed in the system
 * add automake option dist-bzip2
 * use tar.bz2 for rpm build
 * add dist-bzip for the latter
 * use zziplib-lib0.10 package which provides zziplib and libzzip0
 * forgot to add *.la files into the rpm devel package
 * AC_SET_RELEASEINFO_VERSIONINFO => AX_SET_VERSION_INFO
 * adjust all RELEASEINFO -> RELEASE_INFO accordingly
 * and choose 0.10.82.pre1 as a vesion number for testing
	which is possible now as the new macro does it right.
 * forgot override of default_includes in test/*.am
 * somehow ax_prefix gets the wrong input, i.e. _config.h instead of config.h
	so we make that one explicit now

2003-05-13
 * the off64_t mode renaming for 64on32 was done incorrectly and
	it lead to problems on solaris8 - fix it.
 * fixup `make check` to work on subdir build

2003-05-11
 * adapt README file to clarify LICENSE details as suggested by
   a discussion with ACE/TAO developers that want to use the
   library for some compressed xml archives.
 * fix a few zzip_char_t bugs that got visible in MSVC mode as
   reported by Olge Ryabukha
 * kill sfnet strand of the rpm build, the docs are installed
   into the share/groups part and the omf file points to that
   place. The index.html file is the one at sourceforge while
   the other one is the entry point of the local documentation.

2003-04-21
 * quite a few updates are need to --enable-builddir by default
 * watch out: subdir build is the default now!

2003-04-20
 * pick up AX_ENABLE_BUILDDIR
 * throw out ac_subst CONFIG_ARGS
 * use AX_SPEC_PACKAGE_VERSION

2003-04-20
 * memory leak in zzip/file.c - adding free(dir->realname) zzip/zip.c
	in function zzip_dir_close called via zzip_file_close
   (bug reported by Olge Ryabukha, thanks!)
 * "/" is hardcoded for finding subpaths but on windows that could be
	a backslash as well. That should be handled better, we add some
	alternative implementations for dirsep_strrchr and dirsep_casecmp
	but the default is OFF!! even on WINDOWS !!
 * thereby seen another problem: we did deliberately use strcasecmp but
	this symbol is not available on all unix systems - the manual
	says its a BSD4.3 addition. Some other systems call it strcmpi
	or stricmp instead, or they do not have it at all. Here we add
	a configure rule to check for that symbol, and when we do not
	have it then we _enable_ dirsep_casecmp automatically.

2003-01-16
 * change extension of intermediate docbookfiles to ".xml"
 * throw out attempts to compile docs with php

2003-01-15
 * lots of doc updates in between. (forgot to document them).
 * a zzip-config update, and installing an aclocal macro now.

2003-01-06
 * defattr(root,root) !! - why that had been forgotten? hmmm...
 * allowing for --disable-debug to clean cflags from -g entries
	but it has not effect on callframe generation... actually.

2003-01-05
 * we need to stop the manpages.ar bloat - now!

2003-01-05
 * add largefile tests in configure.ac
 * add largefile detection in zzip/conf.h
 * add renamings for zzip_telldir/seekdir in zzip/zzip.h
 * add extra exports in zzip/dir.c to have both 32bit and 64bit names exported
	i.e. zzip_telldir _and_ zzip_telldir64
 * add renamings in zzip/plugin.h to ensure that 64bit-off_t-compiled zziplib
	is not given io-callbacks being from another program that is actually
	compiled as 32bit-off_t and where therefore arguments do not match.
 * detected a dubious export of "aligned4()" - rename this export now.
 * also rename everything having "_io" in the name - just to guard ourselves
	from the case where two -lzzip are in the libpath and one is
	compiled 64bit-off_t while the other is not.
 * rename the younger function s/zzip_open_shared/zzip_open_shared_io/
 * make up functions for zzip_open_ext_io and zzip_opendir_ext_io for the
	case of 32/64bit dual systems where lib is compiled in largefile mode.
	Let the call succeed unless an "io" structure was given - return
	the EOVERFLOW errno as specified in the large.file WG documents.
 * make all parallel defines for 32/64 bit dual off_t dependent on
	defined EOVERFLOW and defined PIC - i.e. only for dynamic linkage.
 * adding LARGEFILE_CFLAGS into zziplib.pc and bins/zzip-config.in
 * NOTE: the default for --with-largefile is still OFF !!
 * NOTE: our zzip_readdir is not depenent on off_t size
 * adding -64 into the --release info for the sharedlibrary build
	it should have been a --variant but such does not exist so far
	in libtool, sorry.
 * extract AC_SYS_LARGEFILE_SENSITIVE and put it into an extra ac-macro
 * clean zzip/file.h from compatibility-install as zzip-file.h
	the structure in there contains zzip_off_t items and they
	should be internal always anyway - so disallow access for
	old programs now just to be sure.
 * adding zzip64 compilation into zziplib.spec - an rpm package will now
	install both 32bit off_t and 64bit off_t variants of the sharedlib
 * note: we do not copy 64bit zzcat into the rpm, perhaps we should?

2002-12-31
 * add make check-sfx
 * add zzipself.c and zzipsetstub.c for the sfx test
 * zzip_file_open takes o_modes flags, clarify in zzip.h

2002-12-27
 * adding scrollkeeper omf file generation in docs/
 * adding omf file installation in docs/make-install-doc
 * replace %post-doc message in zziplib.spec with scrollkeeper-update call.
 * need *.pc *.omf in CLEANFILES now
 * mingw32 crossgcc tests for win32 dirent emulation - did need some fixing

2002-12-23
 * adding generation/installation of pkgconfig files
 * making SDL a build-requirement for rpm packaging

2002-12-22
 * add compatibility subdirectory "zziplib"
 * zzipback.sed to create zziplib/* files from current release files
 * include into distribution tarball
 * IMPORTANT: flashback - 12.24 is now called 10.74 -
	the old source line will cease to exist, only the compatibility
	directory will survive a while - that should put some pressure
	on developers to update to the new source tree. And it saves me
	some work to keep two trees in sync.

2002-12-20
 * fix some problems of make-doc.pl with the current project
 * make a standard inlude info referring to zziplib.h as well
 * unpack htmpages.ar on SFNET and register it on the frontpage
 * don't forget to modify includes of internal headers
 * ...
 * fixed an annoying bug in make-doc.pl
 * include htmpages.ar into the tarball

2002-12-19 Guido Drhaeim <guidod@gmx.de>
 * completed new directory structure - the library is called 0.12.22 as
	to match 0.10.72, it's the same C source in its content
 * install-headers modified - the new structure is now compatible both
	as a shared library (binary compatible) and for the include headers
	too - source compatible not only for function names but also for
	the #include lines
 * 0.12.23 fixes a bug in make doc: the xmlto html was never done.
 * and we implant make-doc.pl which was founded on the base of the local
	zzip-doc.pl but has since gone a lot further. It's currently just
	a third variant but it is now used as the primary master default.

2002-12-18 Guido Draheim <guidod@gmx.de>
 * man page update for zzip_open_shared (it turns out to be a good name).
 * let frontpage point to http://www.gzip.org/zlib for primary zlib site.

2002-12-17 Guido Draheim <guidod@gmx.de>
 * introduce zzip_open_file that may have an extra argument of a factory
	file to share the zzip internal directory if that is possible.
 * introduce zzip_freopen to take an extra argument. The supplied handle
	is freed unless of course one puts ZZIP_FACTORY into o_modes so
	that it becomes a wrapper around zzip_open_file
 * the zzip_open_file, in that case, is the old zzip_open_ext_io call
	but extended with checking the factory stream for a matching
	basename.
 * while being there: introduce ZZIP_ALLOWREAL o_mode that will make it
	that a real file is open/read/close with the standard posixish
	routines. This is most useful when having an obfuscated zip
	file around but during development one wants to sometimes check
	with a real file - so the call mode is something in the middle
	of ZZIP_ALLOWREAL - ZZIP_PREFERZIP|ZZIP_ALLOWREAL - ZZIP_ONLYZIP
	where the middle is for "yet another file not yet included",
	hmm, perhaps a cheat file? ;-)=)
 * cleaning rwops example: fix the mode check, it's not needful to start
	with "r" to be a readfile mode, it may be somewhere in between.
 * introduced __zzip_dir_parse as a cut off from zzip_dir_fdopen_ext_io,
	but it did turn out to be not necessary, so it is made static in
	the end. Still, it works as documenting the control flow better.

2002-12-17 Guido Draheim <guidod@gmx.de>
 * win32 mmap removing extra braces
 * win32 dirent implementation added - new file __dirent.h
	and modified zzip-dir.c
 * a couple of arg[n] -> argv[argn] cleanups.
 * a missing argc > 1 check added.
 * all changes due tests / bug reports by Thomas-dot-Eder-at-nmi-AT

2002-12-16 Guido Draheim <guidod@gmx.de>
 * the infozip `zip` tool puts out messages per default on some machines
	fix by adding >/dev/null in some places of Makefile.am
 * the hpux10.20 platform has sys/mman.h but no MAP_FAILED
	the `man mmap` says it returnes "-1" as done on all other unix
	compatible platforms, so we add a default define ((void*)(-1))
 * update TODO file with some hints on test runs.
 * add --with-zlib configure call to make it easier to specify a path
 * make zlib error messages to point to this configure option.
 * beautify error output on all other options as well.
 * create `make brute` to make a  bruteforce test
 * solaris8 make has a problem with $< ... change sometimes into $?
 * freebsd/darwin have 64bit off_t by default - and that reveals a problem
	in the zzip_telldir/zzip_seekdir code. Let's change the code to make
	it return/receive a byte-offset of "dir->hdr" to "dir->hdr0".
	This makes printing of those values also more intelligible (if ever).
 * ia64-linux reveals a problem in "default_io" since size_t is 64bit.
	We modify --with-wrapwrap to autodetect this case. This is okay
	since zip-files can not get bigger than 32bit offsets anyway.

2002-12-16 Guido Draheim <guidod@gmx.de>
 * merging corrected README.SDL and __hints.h
	courtesy of Thomas-dot-Eder-at-nmi-AT
 * update CFLAGS ac-macros to the new generation as shipped with
	latest ac-archive-0.5.39 (did upload just yesterday...)

2002-12-14 Guido Draheim <guidod@gmx.de>
 * many hours of debugging - I don't remember which flaw came in through
	the latest changes or which others were there before *sigh*
 * e.g. the error-code mappings were refined on linux errno codes
 * fix "caseless" bugs reported by Thomas-dot-Eder-at-nmi-AT
 * fix MSVC related flaw when taking a casted value as an lvalue
 * replace AC_CREATE_PREFIX_CONFIG_H with new style AX_PREFIX_CONFIG_H
 * invent zzip-msvc.sed to turn config.h.in into zzip-msvc.in and
	finalize a fresh zzip-msvc.h via AX_PREFIX_CONFIG_H

2002-12-13 Guido Draheim <guidod@gmx.de>
 * shown by Mike Nordell tamlin-at-algonet-SE we can support win32 mmap
 * modify configure.ac : if have:winbase.h then auto -enable-mmap
 * refactor mmap'ing - it goes into __mmap.h and uses the io->use_mmap value
	and that include file contains both posix and win32 mmap variants
 * rewrite pagesize/mapoff calculations - this should be faster in itself
	and also easier to maintain in the future for being more obvious.
 * rewrite read/write defines and wrapwrap's. The zzip-conf.h does now
	contain defines for _zzip_read/_zzip_lseek - and it should us get
	rid of the nasty win32 problems with redefining _read/read
 * refactor DBG defines - put them into __debug.h - oops, xbuf() was
	an exported symbol, better be not. Use only newer NOTE/WARN macros!
 * implant __hints.h include - that's taken from another project but it
	is so damn useful ;-)

2002-12-12 Guido Draheim <guidod@gmx.de>
 * modifizied zzip-zip.c to include a sanity check on getpagesize ()
	for mmap () functionality and ZZIP_BUFSIZ (the default is good).
 * modify configure.ac : const char strings
 * modify configure.ac : strict prototypes
 * modify configure.ac : all warnings
 * ... and so it reveals something missed along ;-)

2002-08-27 Guido Draheim <guidod@gmx.de>
 * modified zzip-stdint.h along the example patch for FreeBSD done
	by wjpalentstijn at sourceforge in projet uwadv.

2002-07-24 Guido Draheim <guidod@gmx.de>
 * did teach myself php over converting zzip-doc.pl into zzip-doc.php
 * added php-test and another way to create the docbook file, with php

2002-07-20 Guido Draheim <guidod@gmx.de>
 * lots of fixes around docbook/manpage - and generate a zziplib.3
	overview page. Fixes also lots of comments in the sources.
 * modified fopen-mode interpration nicely.
 * added zzip-api for full overview of the zzip's API

2002-07-19 Guido Draheim <guidod@gmx.de>
 * greatly enhanced docbook/manpage generation - including to attach
	author and license info from the respective source file header

2002-07-18 Guido Draheim <guidod@gmx.de>
 * extended the perlscript to generate zziplib.docbook
 * use xmlto to generated manpages.ar
 * install-man3 target and push into rpm-devel package

2002-07-17 Guido Draheim <guidod@gmx.de>
 * no new code, just doc updates. The perl-script is now capable to
	combine _ext_io functions with their non-ext/io cousin, and the
	sources have been modified to honour this new mode.
 * the perl-script is now ready to spit out other data like docbook
	files that we can generate man-pages from later.

2002-07-16 Guido Draheim <guidod@gmx.de>
 * extern-C in SDL_rwops_zzip.h ... and ifdef'able DECLSPEC for staticlinking
 * zzip_fopen allows to greatly simplify SDL_rwops_zzip.c
 * zzxordir.c zzxorcat.c zzxorcopy.c invented
 * lots of new documentation added
	zzip-extio.html zzip-xor.html zzip-refs.html
 * new checks that test the zzxor examples
 * and found an zzxor bug in the course

2002-07-15 Guido Draheim <guidod@gmx.de>
 * Michael-dot-Fink-at-asamnet-dot-de did some code review while
	going after a specialty he wanted to do with the libs, and therefore..
 * PREFERZIP did never test real file - stupid bug, inverted logic....
 * add ZIPONLY zzipmode - we have plenty of options available... on 32bit
	platforms however, since I want to keep with the access-bits in
	the lower parts.

2002-07-12 Guido Draheim <guidod@gmx.de>
 * watcom-c-10.5 project file added
 * used my old watcom compiler to debug the win32 problems, and it turned
	out to be a mismatch in proccall for zlib - while the default in
	zlib.h inferred WINAPI, it turned out that zlib114dll.zip had in
	fact been compiled with __syscall procframe callmethod. Duh!!!
 * some things had to be rearranged in header files - most classically it
	seems that this compiler does not support __inline in C mode.

2002-07-11 Guido Draheim <guidod@gmx.de>
 * convert all API const types into those with _zzip_const and use new
	types like zzip_char_t and zzip_strings_t throughout the sources.
 * add ZZIP_RDONLY and friends. Do not use fcntl bits anymore, and limit
	use routines to convert them into O_DEFs later one in the next
	version of this lib - here we just declare them to be their O_DEFs
	but code should start to pick up ZZIP_DEFs in the meantime.
 * add zzip_fopen and friends as another type of magic wrappers.
 * modify open_ext_io to use receive separate arguments for the two
	zzip extraflags - use these from now on, the others are just
	for limited backward compatibility.
 * modify all _close functions to return "int".

2002-07-10 Guido Draheim <guidod@gmx.de>
 * Marlin Prowell mbp-at-cadalog-inc-dot-com did hit a problem on FreeBSD
	showing off a config problem - the configure.ac macro for off_t is
	written to be _undefined_ if off_t existed in the system. Instead we
	chose to defined zzip_off_t into a "long" as it is needed on win32,
	and since FreeBSD on LFP32 uses OFF64 it did fall flat on calls to
	lseek - without io-wrap the gcc did cast them correctly, so it worked.
 * the fix: use off_t if zzip_off_t undefined. Add a define zzip_off_t into
	long in conf-msvc.h. Add a configure-option to have wrappers for the
	default lowlevel calls as suggested by Marlin but I choose to have
	them off by default - a bsd ports package may start using it.
 * there have not been reported problems in the last three months where only
	developers knew about the new version. Guess it's time to push this
	tarball and announce it on freshmeat.

2002-04-30 Guido Draheim <guidod@gmx.de>
 * fix typo in zzip-zip.c: __USE_MMAP -> _USE_MMAP
 * if zlib.h not found, make it a fatal exit

2002-04-25 Guido Draheim <guidod@gmx.de>
 * adding a few more `make check` rules for bigger zip files
 * add even negative tests
 * add some spaces in error-messaging in zzdir.c and let return
	an exitcode when some argument files could not be read

2002-04-24 Guido Draheim <guidod@gmx.de>
 * apply patch from Mike Nordell
	changes all "const char * *" into the intended "const char * const *"
	changes some typecasts
	corrects comment and adds some
	and two fixes
 * put the updated MSVC6 files into the project, again Mike Nordell
 * me, add new zzwrap.dsp to DIST in Makefile.am, following Mike's addition
 * me, modify some typecasts to use off_t/size_t where appropriate
 * modified COPYING.ZZIP to be a bit more general than just for ZZIPLIB
	as noted by Tomi Olilla
 * zzipwrap demo_callback, how can we do void* and char* at the same time...

2002-04-23 Guido Draheim <guidod@gmx.de>
 * changed ssize_t into int in zzipwrap-mem for portability reasons
 * fix zzip_dir_fdopen_ext for access to uninitialized io-vtable
	(thanks to Lutz Sammer johns98-at-web-dot-de for testing)

2002-04-22 Guido Draheim <guidod@gmx.de>
 * add MSVC6 project files contributed by Mike Nordell tamlin-at-algonet-dot-SE
 * add plugin_io-patch from Mike Nordell
 	and dump the iowrap-patch from Andreas Shiffler from earlier release
 * append the Mike Nordell's zzobfuscated.c to Makefile.am
 * add a new file COPYING for new less-strictly license
	after Tomi Ollila has transferred full copyright to me
 * remove Tomi Ollila from all copyright entries in the sources
	but honour his name where appropriate
 * modify LGPL hints in sources and point to COPYING.ZZIP as well
        and change copyright year info
 * create COPYING.ZZIP with some general LGPL exceptions.
 * use newer AC_CREATE_CONFIG_H macro and start using acinclude.m4
	to allow retooling of the project.
 * and use zzip_off_t instead of ZZIP_off_t (but keep backward compatibility)
 * apply patch for the problem detected and corrected by
	 Steve Dillon steved-at-phone-IVR-dot-com
 * try with an updated ac_create_prefix_config_h due to a bug report by
	John W. O'Brien john-at-jugglers-dot.com
   where his shell-and-echo combination did interpret \\1 twice.
 * add use_mmap flag to plugin_io_t, so the feature can be used
	along USE_MMAP configured setups
 * change USE_MMAP #ifdefs to use if()s and to rely on the
	compiler to do code removal for unreachable code
 * remove tell() part of plugin_io and rename seek()-indirection
	to seeks()-indirection using a tells()-macro in zzip-file
 * also rename s/seek/seeks/ in zzip-zip and kill the (*function)(x)
	for the indirection functions into just function(x)-calls
 * fix zzobfuscated example code - there's no read() without unistd.h
	in unix, and global zzip_install_io had been removed lately.
 * add zzipwrap-io.c and zzipwrap.c from Andreas Schiffler's patch
	on the new ground of Mike Nordell's plugin_io
 * modify zzipwrap sources to work with plugin-io
 * add io pointer to dir-structure, and use it as the default for file-structs
 * __zzip_open_zip is called __zzip_try_open now and takes two arguments
 * typedef struct zzip_plugin_io * zzip_plugin_io_t;
 * zzip_opendir_ext_io .. for magic convenience
 * create new zzipwrap.la in Makefile.am
 * move plugin_io struct from zziplib.h to zzip-io.h
	along with its helper functions
 * change all filesize into zzip_off_t
 * s/zzip_find_disk_trailer/__zzip_find_disk_trailer/ (additional io arg)
 * s/zzip_parse_root_directory/__zzip_parse_root_directory/ (ditto)

2001-10-11 Guido Draheim <guidod@gmx.de>
 * added zzipwrap feature contributed by Andreas Schiffler

2001-09-18 Guido Draheim <guidod@gmx.de>
 * fix bug for `zzip-config --cflags`
	which did print /xx/include instead of -I/xx/include
	(thanks to Roger Ostrander <denor(at)yahoo.com>)
 * change the for-loop in zzip-config - the "for i ; do" is not portable
	because I had problems on solaris-sh. Use "$@" which expands
	to a list (!!) of argv entries for this for-loop even for
	arguments that contain spaces. fascinating.
 * use AC_HELP_STRING ... the `configure --help` screen got a bit
	misformatted lately.
 * add a `make check` rule.

2001-08-19 Guido Draheim <guidod@gmx.de>
 * add that /programs magic to the $prefix check in configure.ac
 * add that /usr/share magic to the $prefix check in configure.ac
 * fix zzip-sdl-rwops.html layout (and copying.zlib link).
 * fix some flaws in sdl-rwops example source (just by code review).
 * add unistd.h include to zziptest.c to silence off the warning about sleep
 * add non-dll compile to extern in zzip-conf.h - does that fix anything?

2001-08-18 Guido Draheim <guidod@gmx.de>
 * SDL_rwops_zzip.c/SDL_rwops_zzip.h SDL-example file.
 * change all filename args to be const
 * add ZZIP_const to zzip-conf.h, but do not use it for now.
	It seems that all current lib-users have a modern compiler...

2001-08-16 Guido Draheim <guidod@gmx.de>
 * export zzip_dir_free / zzip_dir_alloc
 * bring in fileext field to check for multiple fileextensions.
 * make zzip_open check for multiple (default) fileextensions.
 * add ifdef USE_ZZIPLIKES to implement feature proposed by
	Julien Cayzac <deepmind(at)jessica.brainlex.com>
 * make zzip_open check along the dirpath for a zip-file
	a feature proposed by Lutz Sammer <Johns98(at)web.de>

2001-08-15 Guido Draheim <guidod@gmx.de>
 * add "stroustrup" style to all c files to ensure that casual
	developers do not trash the layout unintentionally.
 * modify examples a bit with more newlines / comments.
 * re-license examples to ZLIB license
 * include COPYING.ZLIB into doc_FILES

2001-08-10 Guido Draheim <guidod@gmx.de>
 * change frontpage of webpages - including a few crosslinks to
	other projects

2001-08-07 Guido Draheim <guidod@gmx.de>
 * split version X.Y.Z into -relese X -version-info Y:Z
 * adding --disable-thread-safe, and add -thread-safe linkage by default
 * modify releaseinfo for with-debug
 * install-zip target to make a windows package
 * change serial to _vendor - rpm has never been used for other than linux

2001-08-06 Guido Draheim <guidod@gmx.de>
 * reorganize docs, update docs to new style
 * split version X.Y.Z into -release X.Y -version-info Z

2001-08-05 Guido Draheim <guidod@gmx.de>
 * add zzip-msvc.h, a header file that will be included when
	a windows compiler is detected in zzip-conf.h, e.g.
	_MSC_VER or __BORLANDC__
 * mv index.html zzip-index.html and add makerule that patches
	the VERSION in before installing it as index.html
 * add a makerule to build sourceforge htdocs and add it
	to the rpm build.
 * add generic patches for libtool into configure.ac

2001-07-18 Guido Draheim <guidod@gmx.de>
 * check up bugs in automake 1.4beef, now using a rule to
	override AUTO* in non-maintainer-mode. we also set
	libtool to silent in non-maintainer-mode
 * docs are installed to $datadir/doc/$package as default
 * rpms can now be built.
 * create --with-docdir and let it be filled in rpm-configure
	so that docs will end up where rpmopt says it should.
 * ChangeLog goes into %doc of rpms.
 * uhmm, forgot to install the test*.c files into the docdir
 * the bin rpms had not been built
 * compress into base and devel package
 * errnolist should be static

2001-07-17 Guido Draheim <guidod@gmx.de>
 * nuke Makerule system, use plain autotools style including
	configure && make install to rebuild
 * update Makefile.am (which is new now)
 * use next-generation autotools as a basis, update configure.ac
	to newer style, update configscreen at the end of file
 * check perl in configure and let it default to "echo"
 * check HAVE_ defines coming from config.h - it is not anymore
	included by zzip-config.h, so each config-def usage
	must now use the prefix'd version
 * provide a zziplib.spec file

2001-01-05 Guido Draheim <guidod@gmx.de>
 * apply zzip_seek-patch from Ted [sourceforge.net/users/tam4]
 * -> 0.10.12

2001-01-02 Guido Draheim <guidod@gmx.de>
 * gone through a couple of trial and error versions until
	I finally arrived at a mingw32msvc zzip.dll - basically
	I had to update the included libtool files to 1.3.5
	and the zzip.dll source-files have a ZZIP_DLL define
	before the first include that points to zzip-conf.h -
	and there is the magic about dllexport and dllimport.
	Hopefully the same can run under MSVC without further
	changes to the sources.
 * try to package a zip-file that contains the relevant files
	from the mingw32msvc install, so that windows user
	get a clean dll without the need to compile it.
 * -> 0.10.11

2001-01-01 Guido Draheim <guidod@gmx.de>
 * received a patch for MSVC compatibility from Ted A. (@lehigh.edu)
   which made me to look after the project again. There
   were two support requrest during summer that asked for
   MSVC support, especially a prebuilt dll and headers.
 * added AC_PREFIX_CONFIG_H that I made originally for
   another project. It enables me to push the configure
   detections into a global include/-dir. Ted's patch
   used an `IFDEF MSVC DEFINE HAVE_SYS_TYPES_H` which
   is a symptom of using config.h&HAVE_CONFIG_H and
   zzip-file.h&IFDEF HAVE_xx. Change all header-IFDEF-HAVE
   to use IFDEF-ZZIP_HAVE, and install zzip-config.h
   along with the other headers, change the original
   include"config.h" in zzip-stdint.h and make it rely
   on zzip-conf.h - where zzip-conf.h includes zzip-config.h
   and lists the detected features to be used, along with most
   important extra defs for MSVC compatibility. (!!!)
 * added AC_COMPILE_CHECK_SIZEOF (cross-compile clean!)
   to always use the correct STDINT defines. Delete
   use of glib.h if present, it's not needed anymore.
 * use ZZIP_off_t instead of off_t
 * prepare for a MSVC' ZZIP.DLL with _zzip_export defines..
   -DZZIP_DLL -> declspec(dllexport)
   -DHAVE_LIBZZIP -> declspec(dllimport)
 * make sure, __attribute__ is okay with non-gcc compilers
   and attache packed-attribute to *all* zipformat-structs. The
   whole of zzipformat.h is also enclosed in pack-pragmas.
 * -> 0.10.7
 * create zzip-config.in and install it.
 * mingw32msvc has shown that we need -lz on the command-line
   to resolve a few symbols
 * go automake
 * install zzip-config along
 * mingw32msvc does not build with cross-make.sh as seen on
   http://libsdl.org/Xmingw32 - it needs the cross-tool/lib
   just as well. There we have LDLIBADD in Makefile.mk now.
   Now everything works for making win32 target with cross-tools.
 * The rest of the make-struture should ensure to be able to
   create an MSVC nmake Makefile - the Makerule.am is clean and
   can be directly included... hopefully.
 * -> 0.10.8


2000-06-03 Guido Draheim <guidod@gmx.de>
 * Tomi did a final test round for solaris, oh well,
   configure needs some fixes...
 * checking for myself at the university's solaris2.6 with
   gcc-2.95.2, hmm... there's an sys/int_types.h that
   seems to define some of C9X int-types, put it into
   the configure.in and know about it in zzip-stdint.h
   (btw, glibc seems to have include/inttypes.h that
    includes stdint.h).
 * tomi did report a problem with gcc 2.7.2... probably
   sth. about -rpath
 * release 0.10.6 and going to announce it...

2000-05-31 Guido Draheim <guidod@gmx.de>
 * created the main doc html files, index.html, zzip-zip.html
   and zzip-file.html. the README and README.API is deleted,
   and the TODO is a bit updated.
 * This version is meant to be announced very soon now.
 * release 0.10.5 and contact Tomi for his final nod.

2000-05-31 Guido Draheim <guidod@gmx.de>
 * ifdef HAVE_SYS_STAT_H where stat is used. Removed
   the stats from zzip-file.h - it will now just try
   to open the file or zip-file and assume that
   open(2) will fail if the file is not there. I do
   hereby assume that the returned fd is seekable..
 * two new functions, zzip_dir_real and zzip_file_real
   to check if the ZZIP_XX structures are wrapping a
   real file or directory. Otherwise it is a zip-archive
   or a zip-contained file respectively.
 * zzip_compr_str does now know about stat-types and
   will return names for S_ISTYPE bits.
 * remove zzip_fd. It was identical to zzip_dirfd
   whose name is consistent with dirent.h
 * zzip_stat -> zzip_dir_stat since it does not neither
   return a stat structure nor does it use stat(2)
 * Checking the doc-comments

2000-05-30 Guido Draheim <guidod@gmx.de>
 * zzip_fp -> zzip_file and _fp_ functions to _file_
 * zzip-stdint.h to ensure iso c9x typedefs
 * ZZIP_GET16/32 macros in zzip.h header.
 * minor updates in cosmetics and documentation
 * have full replacements for dirent.h functions,
   the zip-specific functions are now called
   zip_dir_open (and more alike). The zzdir can
   now show a list of files in a real directory,
   or magically the files contained in a zip-arch.
 * release 0.10.4

2000-05-29 Guido Draheim <guidod@gmx.de>
 * zzipmain -> zziptest and some cleanups
   in the Makefile. zziptest is now just
   another test program, so it gets compiled
   in full similarity with zzcat and zzdir.
 * removed zziptest from install-bin
 * the long forgotten simplicistic dependency
   rule is now in here too.
 * cosmetics to pointer declar' stars in
   zziplib.h
 * ZZIP_FP -> ZZIP_FILE and
   zzip-fp.h -> zzip-file.h
 * removed the use of zzip.h where suggested
   by Tomi. (use public zziplib.h instead)
 * removed errno.h from zziplib.h
 * printing configuration at end of configure.
 * add install-includeHEADERS rule, so that
   zziplib.h does now get installed.


2000-05-27 Guido Draheim <guidod@gmx.de>
 * zzipformat.h carries information about the
   zip-file structure. The typedefs are now
   used to access members of the zip-archive
   (instead of using the offsets directly).
 * notice that the members are declared
   bytewise due to compiler semantics that is
   making a structure packed&aligned to the
   greatest common size (which would be 4).
   setting as attribute(packed) is not very
   portable.
 * reworked the files, esp. zzip-zip.c and
   zzip-ar.c
 * _USE_MMAP in autoconf, including option
   and CFLAGS setting. instead of open/read
   to a scan buffer the zip-file is mmapped
   and scanned directly.
 * _LOWSTK in autoconf, including option and
    a CFLAGS setting. Still need a scan buffer
    and this flag tells us not to carve the
    scan buffer from the auto-stack and use
    instead some malloc for the scan buffer.
 * DEBUG in autoconf, including option and
    a CFLAGS setting. The macros should now
    be able to be found everyone and only.
 * revamped EDH find routine, dumped old-style.
 * revamped read-EDH to use mmap and to use
   copy/read only for the parts we need. The
   read-EDH buffer is not limited to 32K
   anymore and can have any length. No realloc
   need anymore.
 * opendir does not allocate the cache.buf32k
   anymore - it does have no need for it. the
   buffer is still cached for subsequent Open
   calls. May want to setvbuf later.
 * a few renamings to be more bits/dirent.h-like,
   so many things are now called d_name and
   d_reclen instead of name or next as before.
 * renamed the package to "zziplib" as being
   suggested by Tomi.
 * release 0.10.3

2000-05-26 Guido Draheim <guidod@gmx.de>
 * -release at link-line for the lib
 * zzcat.c as a test program, which uncovered
     a few flaws.
 * a flaw w/_fd_opendir when being given a
     non-inited err-return-variable and no
     error did occur.
 * the error to errno table was incomplete,
   and the ZZIP_INFLATE error is superfluous.
 * zzdir.c as a test program, needs still to
   be given the full zip-arch name.
 * changed member names of zzip_dirent to
     match similar names in dirent and stat,
     (ie. d_name and st_size) and added furthermore
     a z_size that returns the compressed size.
 * little updates to names and comments.
 * release 0.10.2

2000-05-22 Guido Draheim <guidod@gmx.de>
 * merge with zip08x-0.9.6 - actually just
   a bugfix for zzip_fp_open IGNOREPATH -
   I even applied it in a different way than
   Tomi did.
 * bugfixes for zzip_open, it did not always
   set errno(2) to the value of zzip_errno.
 * The zzip_read is now separate from
   zzip_fp_read, so that setting errno(2) is
   only done in zzip_read, not in the zzip-touching
   function.
 * zzip-doc.pl to generate libzzip.html
 * release 0.10.1

2000-05-21 Guido Draheim <guidod@gmx.de>
 * drop-in replacements ready for daily use,
   just use zzip_open with a normal path to
   a fs file. If that file is not found in
   a normal fs directory, then the path is
   splitted into a basepath and a filename,
   a ".zip" is appended to the basepath to
   form a zip-arch name, and the filename
   is then searched inside the zip-arch if
   that one exists. Very nice for distribution
   of a subdirectory full of scripts and that
   is what it is meant for - the PFE will be
   able to use this one for Forth-files.
 * have a VERSION file containing "0.10.0" and
   add some rules to Mk-rules
 * start with in-source docs that will be used
   for documenting the lib calls. A jdoc like
   doc extractor should be able to grok these,
   I'll see to release docs with the next
   release, but this time...
 * make it release 0.10.0 and upload it to
   pfe.sourceforge.net/pub where it will lurk
   around until I contacted Tomi. He seems to
   be currently working on zip08x

2000-05-17 Guido Draheim <guidod@gmx.de>
 * decided to have it actually called zzip_ and so
   all of the stuff is going to get renamed again.

2000-05-16 Guido Draheim <guidod@gmx.de>
 * obtained Tomi's code and made changes so that the
   resulting library is more portable, esp. the glib
   usage is being abolished. The name of the library
   (and package) has been renamed to libunzip and
   all types, consts and funcs to now carry the unzip_
   prefix.
 * autoconf !!!!
 * drop-in functions that can be made to easily replace
   dirent and filefd functions.


Mon Apr  5 11:23:46 2000  Tomi Ollila  (too@iki.fi)

	* Release 0.9.5

	* Made some tiny modifications after Matthew's code.

Sun Apr  4  8:30:00 2000  Matthew D. Campbell (matt@campbellhome.dhs.org)

  * Cleaned up compile warnings in SuSE Linux

  * Cleaned up code for cross-compilation with Xmingw32

  * Modified zipx.h to support linking to C++ apps

Wed Jul 28 11:23:46 1999  Tomi Ollila  (too@iki.fi)

	* Release 0.9.4

	* Fixed support for simultaneous open files with each
	  open `ZipFile'.

Thu Jul 22 11:57:46 1999  Tomi Ollila  (too@iki.fi)

	* Release 0.9.3

	* Added premilinary README.API.

Sat Jul 10 11:49:58 1999  Tomi Ollila  (too@iki.fi)

	* Release 0.9.2

	* Fixed code to read start-of-file offset from local header
	  instead relying directory info is sufficient (were TOO naive
	  about that, the first user reported this library won't work
	  with his archives).

Sun Jun 13 20:40:11 1999  Tomi Ollila  (too@iki.fi)

	* Initial release (0.9.1).

guidod@gmx.de 2018-02-05