History of changes for DeuTex

DeuTex 4.4.902 (AYM 2005-08-31)

• Build: vanilla make compatibility: removed all occurrences of "$<" not in inference rules in the makefile.
• Build: vanilla make compatibility: removed all "\" followed by empty lines. On NCR MP-RAS, make quotes all empty lines that follow !
• Build: there is now a "./configure" script (GNU autoconf work-alike).

  Installing somewhere else than /usr/local is now done by giving ./configure the --prefix option.

  CC, CFLAGS and LDFLAGS are now auto-detected. The defaults can be overridden by passing ./configure the --cc, --cflags and --ldflags options.

  The fixed-size types (Int16 etc.) are now correctly set on all LP64 platforms, not just Alpha.

• Build: the OPTIONS section is now updated automatically. This impacts no-one but hackers.
• Code: two char[8] were used as char[9] (caught by NCR/MetaWare cc on NCR MP-RAS).
• Code: C89 compatibility: removed a few "//" comments that had slipped into the source.
• Code: a few functions were declared as static but defined as extern (caught by NCR/MetaWare cc on NCR MP-RAS).
• Code: static- and const-correctness fixes.
• Code: removed the little-used Legal() function.
• Doc: man page: in DESCRIPTION, added a short list of examples to help new users getting started. Added a FILES section. Made the SEE ALSO section a bit more specific.
• Doc: in --help and the man page, divided the options into groups and sorted them alphabetically within each group. It seems to be clearer that way.
• Doc: added a FAQ.
• Doc: obfuscated all email addresses (tr .@ +=).
• Game: Strife: the SCRIPTnn lumps are now extracted in human-readable form into the scripts/ directory. The reverse operation is not implemented. Since we understand the Strife script format only partially, the output format is still in flux. New option -scripts to extract only the scripts. The reverse engineering was done in collaboration with Matthew W. Miller.
• Graphics: it's now possible to build wads that use a custom PLAYPAL. Previously, DeuTex always used the PLAYPAL from the iwad when building. It now uses the one in lumps/ if it exists. Thanks to Ras2 for reporting the problem.
• Graphics: rewrote COLdiff() as a macro. On my system, the time to build a wad made of a single 1M-pixel patch is down from 1.32s to 0.85s (35% faster), which is not surprising since DeuTex used to spend about 40% of its time in COLdiff(). The time to rebuild the Strife iwad is down from 8.6s to 7.2s (16% faster).
• Graphics: removed the "quantisation is slow" warnings that DeuTex used to spew when composing from PPMs and 24-bit BMPs. Those warnings were relevant to an old quantisation algorithm that has not been used since at least version 3.8 (#ifdef QUANTSLOW).
• Graphics: lifted the arbitrary limit of 256 patches per texture. DeuTex will now accept as many patches as the wad format allows (32,767). If a texture definition has more than that many patches, DeuTex will discard the excess patches with a warning, instead of dying with a cryptic message like "Line n: Illegal char '*'".
• Graphics: lifted the arbitrary limit of 4096 on patch width. DeuTex now handles patches as wide as the wad format allows (32,767).
• Graphics: textures wider than 4096 are now accepted with a warning. PrBoom 2.2.3 is known to take textures as wide as 16,384 pixels. 8192×128 textures make XDoom 20001001 crash in Z_Malloc. 1024×128 textures make Doom freeze in R_Init. The maximum acceptable widths might be higher if the height of the texture is lesser than 128 but I haven't looked into it. Thanks to David Damerell for testing.
• Misc: removed the annoying startup banner.
• Misc: if no command is given, emit a meaningful error message instead of suggesting to switch to WinTex. And exit with code 1 instead of 255.
• Misc: all messages now look like this :

  c code string

  c is the class of the message : "i" for information, "w" for warning, "E" for error, "B" for bug.

  code is a unique 4-character alphanumeric code which unambiguously identifies the message.

  Some of the messages have been rewritten to be more informative (mention the filename, the nature of the error, etc.).

• Misc: at the request of Kim "Sparky" Parrott, all messages are now copied to a file named "deutex.log" ("deusf.log" for DeuSF). The default log file name can be overridden through the "-log" option. If you don't want a log file, try "-log /dev/null" (Unix) or "-log nul" (DOS).

  The log is only written if a command that works with wads is given. --help, --version etc. do not create a log.

• Misc: removed the 30 command-line arguments limit.
• Misc: made exit status a little bit more normal (2 for errors and 3 for bugs instead of -5 and -10).
• Sound: new option -rate to specify what DeuTex should do when including sound files whose sample rate is not exactly 11025 Hz. The choices are :

  reject

  consider it a fatal error and exit immediately with a non-zero exit code.

  force

  emit a warning and force it to 11025 Hz by resampling up or down.

  warn

  emit a warning but include it as is anyway.

  accept

  silently include it as is.

  Previously, the default (and only option) used to be "force". It's now "warn".

  Thanks to Matthew W. Miller for telling me about this issue (which he did in 1999 ; the six-year delay is mine, all mine).

• Sound: write errors while extracting PC speaker sounds are now actually detected and reported.

DeuTex 4.4.0 (AYM 2000-01-05)

• Game: Hexen: musics are now identified and extracted properly.

  The old music identification code assumed that any lump whose name does not begin with either "D_" or "MUS_" can't be a music. It worked for Doom, Heretic and Strife but, for Hexen, it caused all musics to be seen as plain lumps and extracted accordingly into the lumps/ directory. DeuTex even tried to interprete STALKR and WINNOWR as pictures and said silly things about them having a "width greater than 4096".

  The new code really checks whether the lump begins with "MUS\x1a" instead of just looking at its name. A lump is now deemed to be a music if and only if it begins with "MUS\x1a".

  As a side-effect, certain operations (appending sprites and flats and merging) must have become slower. Furthermore, since these used to blindly assume that any lump whose name begins with either "D_" or "MUS_" is a music, their semantics might have changed. If you find they don't do what you want, try again using the -musid option and let me know whether it improves your condition.

• Game: Hexen, Doom alpha 0.4/0.5: levels are now properly extracted and included.

  There have been changes in the undocumented details of DeuTex's behaviour with respect to levels. The one that is most likely to be noticed is that, when including a level, DeuTex now copies the entire contents of the levels/ pwad, starting from the level label. Previously, it included at most the 11 following lumps, and only if they had the expected names (THINGS, VERTEXES and so on).

  But, basically, if the levels/ pwads contain, as they should, all the needed lumps and nothing else, there shouldn't be any trouble.

• Game: Heretic and Hexen: does not abort anymore with "Bug: *** idinx (12) ***" when trying to include the graphic lumps (resp. CREDIT, E2END, FINAL1, FINAL2, HELP1, HELP2, TITLE and CREDIT, FINALE1, FINALE2, FINALE3, HELP1, HELP2, INTERPIC, TITLE). More generally, DeuTex now accepts to compose wads even when there are graphic files in lumps/.
• Game: Hexen: does not abort anymore with "Height of FLAT ./flats/x_001.ppm is not 64 or 65" when trying to include flats X_001 through X_011. In addition, DeuTex now just emits a warning instead of aborting for other oddball heights (i.e. not 64, 65 or 128). Have fun. ;-) This is true for all iwads, not only Hexen.
• Graphics: the annoying "quantisation is slow" warnings now appear at most once.
• Misc: to disambiguate the "<count> warnings omitted" message, added optional scope prefix and changed the picture extraction function to use it.
• Misc: got rid of the "don't bother Olivier" banner. People must have got the message by now.

DeuTex 4.3.0 (AYM 1999-12-24)

• Graphics: fixed ancient bug where DeuTex sometimes failed to include custom patches if they were not explicitly listed in the [patches] section. If the first patch used in texture1.txt was a custom patch, it had to be listed in [patches] or DeuTex would forget to include it. This is the same bug Olivier mentioned in the home page :

  The support for wall patches in DeuTex has been modified. You must now explicitely declare all your patches in a [PATCHE] section.

  If you don't do this, DeuTex will attempt to work as usual, but there seems to ba a bug in this part of the code, so sometime some needed patches are not loaded.

  After some summary testing, looks like it's fixed.
• Misc: more error handling improvements.
• Misc: bumped version number and cleaned things up for public release.
• Misc: decreased maximum number of warnings per picture from 10 to 5.
• Platform: fixed ftruncate() being undeclared when compiling with DJGPP and updated the building-on-DOS section of the doc.

DeuTex 4.2.2 (AYM 1999-11-20) (not for public consumption)

• Misc: made certain failure messages more informative.
• Platform: fixed several bugs that showed in the DOS precompiled executables for 4.1.0 and 4.2.0 (most common symptom : DeuTex aborting with a "Can't read WAD" error message).

  Lengthy technical explanation : in 4.1.0, I removed the "huge" pointer qualifiers that were scattered throughout the source not unlike nitrates in groundwater. The reasoning was that, since DeuTex is always compiled in the "huge" memory model anyway, those qualifiers were redundant. As I found out at the end of a long and painful debugging session, they weren't.

  Had I read the doc of the compiler, I would have known that, even when in the huge memory model, pointers are "far" by default, not "huge". Far pointers wrap around at 64 kB ; this is not what you want when you're trying to work with lumps larger than that. And, apparently, there is no way to specify that pointers should be huge by default.

  On top of that, there was a genuine bug in WADRreadBytes2() that would have prevented the DOS port from working, even if all pointers had been huge. But this one was fixed in 4.2.1.

  I switched to DJGPP, with which you can get working executables without having to contaminate your code with carcinogenic keywords. The bad news : firstly, the executables are somewhat larger. Secondly, since DJGPP executables use protected mode, they tend to be more fussy.

  Thanks to Kim Parrott for reporting the bug and alpha testing my fixes.

  All the above applies only to the DOS precompiled executables. Other platforms did not have these problems.

DeuTex 4.2.1 (AYM 1999-11-16) (not for public consumption)

• CLI: fixed segfault on "deutex --vers".
• Graphics: new option -usedidx. When called with this option, DeuTex scans all the graphics in the wad and prints statistics about which palette indices they use. (By "graphics" is meant "any data that is converted into an RGB triplet by looking up PLAYPAL or TITLEPAL". That includes flats, graphics, patches, sneaps, sneats and sprites.) I've added this command for my own use, to help me decide which index should be used to store the transparent colour for Hexen.
• Misc: made certain failure messages more informative.
• Misc: made printing of lump names garbage-proof.
• Platform: fixed a huge DOS bug that made DeuTex fail with a "Can't read WAD" error whenever it had to read more than 65535 bytes from a wad at once.
• Platform: flushing stdout before writing to stderr so that messages come out in the right order when both outputs are redirected.
• Sound: all conditions that used to be fatal errors when extracting sound lumps now just elicit a warning message, indicating which lump it was and what action was taken.

DeuTex 4.2.0 (AYM 1999-11-14)

• Doc: fixed error in documentation of -pkgfx, -pknormal and -usedtex.
• Game: Strife: fixed DeuTex aborting when extracting textures for versions of Strife >= 1.1. The problem was that Strife 1.1 and above use a different format for the TEXTURE1 and TEXTURE2 lumps (Strife 1.0 uses the same format as Doom). New options "-tf strife11", "-itf strife11" and "-otf strife11" to support that format. Option -strife has been changed to imply "-tf strife11". New option -strife10 that is identical to -strife except that it does not imply "-tf strife11". Summary :
  • if you have the Strife 1.0 iwad, use "-strife10" (or "-tf normal"),
  • if you have Strife 1.1 or above, use "-strife" (or "-tf strife11").

  Thanks to Kim Parrott for reporting the bug and Len Pitre for pointing me in the right direction.

• Sound: fixed two bugs in reading Sun audio (.au) files. Fixes error "WAV: can't read data of./sounds/foo.au" (sic) when trying to build a wad. One of these bugs prevented from reading Sun audio files on little-endian machines. It had been there for a long time ; v3.8 has it and the v3.6 binary behaves like it had it too. I doubt that anyone had ever been able to use .au files on little-endian machines before.

DeuTex 4.1.0 (AYM 1999-11-01)

• CLI: new options -sneas, -sneaps and -sneats.
• Code: removed all occurrences of "huge" on the theory that, on platforms where it's meaningful, we always use the huge memory model anyway.
• Code: replaced certain occurrences of "Int32" by "iolen_t".
• Code: replaced certain occurrences of "256" by "NCOLOURS".
• Doom alpha 0.4: AMENA0 and MSKUL* are now correctly recognized as graphics and not as lumps anymore. The 21 graphic lumps that ended up in lumps/ are now properly extracted (into sneaps/ and sneats/). (The first item involved propagating to IDENTgraphic() the changes made to PICtoRAW() in v. 4.0.2. The second item needed heavy hacking, creating a new image type (christened "snea") and managing an alternate palette for TITLEPAL.) Still extracted as lumps : GNUM[0-9] and HUFONT.
• Doom alpha 0.5: the 86 graphic lumps that ended up in lumps/ are now properly extracted (into sneaps/ and sneats/). Still extracted as lump : HUFONT.
• Graphics: errors that used to cause DeuTex to give up on extracting a picture now just make it skip the rest of column. It also prints detailed messages about what it didn't like and in which picture it occurred instead of bailing out silently.
• Misc: new option -di to debug entry identification. Useful mainly to hackers.
• Misc: cosmetic changes in the generated wadinfo.txt and in the phase messages.
• Misc: no more messages "Creating PWAD" and "WAD is complete..." during level extraction.
• Misc: set a limit of 10 warnings per picture, to prevent invalid pictures from uselessly flooding the output.

DeuTex 4.0.3 (AYM 1999-10-02)

• CLI: new option -doom2 as suggested by Matthew Miller.
• Graphics: now accepts to extract pictures as large as 4096×4096 (previously the limit was 320×200). This fixes "Failed to write sprite" errors when trying to extract PSYBA0 and PSYBB0 from strain.wad. Thanks to Matthew miller for reporting the bug.
• Misc: added a useful URL to the GIF warning.
• Platform: now builds without errors on FAT filesystems (replaced .deutex and .deusf by tmp/_deutex and tmp/_deusf).
• Sound: fixed a bug that caused DeuTex to extract sounds with unlikely sample rates like 4 GHz whenever the sample rate in the lump was higher than 32767 Hz (for example DSVILACT and DSVILSIT from ncc1701.wad, with a sample rate of 44.1 kHz). Thanks to Matthew Miller for reporting the bug.

DeuTex 4.0.2 (AYM 1999-09-19)

• CLI: new options
  • -doom02 (implies -ipf alpha, -itf none and -itl none)
  • -doom04 (implies -ipf alpha, -itf nameless and -itl textures)
  • -doom05 (implies -ipf alpha and -itl textures)
  • -doompr (implies -ipf pr)

  Use those options where you would have used -doom and friends. For example, to extract the contents of the Doom 0.4 iwad that is in c:\doom0_4, type "deutex -doom04 c:\doom0_4 -xtract".

• Code: replaced certain unjustified uses of Int32 by long.
• Doc: removed old/readme.txt. It's so out of date that it's more confusing than useful.
• Game: Doom alpha and Doom PR: it's now possible to extract graphics, patches, sprites and textures from those iwads. Three new options :

  -ipf {normal|pr|alpha}

  Use alpha for Doom alpha 0.2, 0.4 and 0.5.
  Use pr for Doom PR (press release pre-beta).
  Use normal for everything else.

  -itf {normal|nameless|none}

  Use none for Doom alpha 0.2.
  Use nameless for Doom alpha 0.4.
  Use normal for everything else, including Doom alpha 0.5.

  -itl {normal|textures|none}

  Use none for Doom alpha 0.2.
  Use textures for Doom alpha 0.4 and 0.5.
  Use normal for everything else, including Doom alpha 0.5.

  You shouldn't ever have to use those options directly. It's better to use just -doom02, -doom04, -doom05 and -doompr, which take care of setting ipf, itf and itl properly for you.

  Note that extracting levels and some other lumps from the Doom alpha iwads does not work yet.

• Platform: new target in the makefile to generate a binary DOS distribution with the executables and the user documentation in DOS format, with DOS-ish names.

DeuTex 4.0.1 (AYM 1999-09-10)

• CLI: reworked the command line arguments parsing, with the following consequences.
  • Options can now be abbreviated freely, as long as the abbreviation is not ambiguous. For example, you can use -heretic, -hereti, -heret, -here or -her but not -he because that could also be the abbreviation for -help (or -hexen, for that matter). On the other hand, -h is allowed because it's not an abbreviation (there's really a -h option).
  • -heretic and -hexen now work (they were "hidden" by -h[elp]).
  • -v@ has been split in -v0, -v1 ... -v5 because the new code does not allow excess characters after an option. -vstring where string is anything else than "0" through "5" now triggers an error (it used to be accepted silently). I hope no one relied on the old undocumented behaviour.
  • Certain silly command line arguments that would have worked before would now trigger an error. For example, it used to be possible to type -extramarital or -extermination for -extract but not anymore. The old code defined relatively short options (-ext) and accepted command line arguments as long as the defined option was an initial substring of the command line argument. The new code does the reverse; it defines relatively long options (-extract) and accepts command line argument as long as they're an initial substring of the defined option.
• Code: replaced direct testing of __MSDOS__, __OS2__, __GNUC__, __BORLANDC__ by DT_CC and DT_OS. This is hopefully going to make Udo's job a bit easier.
• Code: now uses the same fopen() modes for all platforms: "{rw}b" for binary mode and "{rw}" for text mode, as per the ANSI/ISO C standard. This will fix the problem Udo Munk reported with the Cygwin build opening binary files in text mode and thus failing miserably. Note that certain DOS C compilers can be configured so that "{rw}" opens files in binary mode. Don't do that ! If you have problems with text files on DOS, make sure your C compiler is configured so that "{rw}" opens files in text mode.
• Code: added to the distribution archive gifcodec.c that I had forgotten to include (it's not used anyway).
• Code: added to the distribution archive src/{deusf,deusfos,deutex,deutexos}.def that I had forgotten to include. I guess that's Windows/OS/2-only stuff.
• Doc: updated making.txt and renamed it as INSTALL for homogeneity. Removed obsolete reference to alpha.sh and the file itself.
• Doc: made more legal updates.
• Doc: documented DOOMWADDIR in the man page.
• Misc: changed the default graphics format for Unix from GIF to PPM, so that fewer user sites are broken if and when GIF support is removed. For the same reason, added a warning message when -gif is used or the first time an image is read from a GIF file.
• Misc: changed the lookup order for images to PPM, BMP, GIF (was BMP, GIF, PPM).
• Platform: fixed a couple of things that didn't work on 16-bit platforms (real-mode DOS).
• Platform: now compiles on DOS with Borland C++ 4.0.
• Platform: now compiles on DOS with MSC 6.0. The MSC 6.0 build is functional but limited because it can't allocate blocks larger than 64 kB, which is insufficient for certain images. I can't use halloc() instead of malloc() because it does not supporting resizing (i.e. there's no hrealloc() function).
• Platform: in response to Udo's remarks, DJGPP and Cygwin are now properly identified (__DJGPP__ and __CYGWIN__).
• Platform: added sanity checks on specified-size types Int32 and friends.

DeuTex 4.0.0a3 (AYM 1999-09-05)

• Code: removed incongruous #define-ing of O_BINARY and SEEK_SET.
• Code: after Udo Munk's report, fixed warnings in
  • src/color.c(74)
  • src/ident.c(583)
  • src/ident.c(658)
  • src/mkwad.c(78)
  • src/mkwad.c(79)
  • src/mkwad.c(80)
  • src/mkwad.c(81)
  • src/picture.c(903)
  • src/picture.c(912)
• Doc: updated making.txt.
• Legal: as agreed to by Olivier Montanuy, DeuTex is now entirely GPL'd. Well, almost entirely, since it includes code written by different authors in lzw.c and elsewhere. Changed the notices in the source files and added new file LICENSE to clarify things.
• Makefile: should now work with all C compilers (removed -Wall from CFLAGS).
• Makefile: clean now removes the DOS executables if they exist.
• Makefile: does not compile with debug information in by default anymore.
• Makefile: new targets dall, ddt, dds, ddeutex and ddeusf for compiling with debug information and all warnings.
• Makefile: new target help.
• Makefile: new target distdos.
• Platform: replaced unlink() by remove() for portability. Thanks to Udo for reporting this.
• Platform: on 8.3 filesystems, make should not choke on docsrc/changes.html anymore. Thanks to Udo for reporting this.
• Sound: corrected some misleading endianness comments in sound.c.

DeuTex 4.0.0a2 (AYM 1999-08-14)

• Game: easier to use with Strife (now looks for "strife1.wad", new option "-strife").
• Game: easier to use with Hexen (new option "-hexen").
• Command line: new options "-hexen" and "-strife".
• Command line: new option "--version" (prints version number and returns 0).
• Doc: various changes in the man page, in the output of "-help" and "-man" and elsewhere.
• Makefile: various improvements.
• Distribution: set modes straight.

DeuTex 4.0.0a1 (AYM 1999-08-12)

• General: fixed many segfaults that were caused by attempts to fclose (NULL).
• Game: Strife is now supported.
• Wad: new options -be, -le, -ibe, -ile, -obe and -ole to control the endianness of the wads. Caution: don't use them if you don't know what you're doing ! As far as I know, wads are always little-endian regardless of the architecture of the host. Therefore, I see no reason for someone in his/her right mind to create a big-endian wad. Those options are here more for the sake of completeness than anything else.
• Wad: made "%" legal in names, to deal with Strife's "INVFONG%" and "INVFONY%".
• Wad, graphics: end-of-flats marker is now F_END by default instead of FF_END. The reason for this change is that, with F_END, you don't need DeuSF to get Doom to see your new flats. Should you need to, it's still possible to get FF_END by using -fend.
• Graphics: the default transparent colour is now a dark blue-green (rgb:00/2f/2f). It used to be cyan (rgb:00/ff/ff) which was blindingly bright, especially compared to the usually dark colours used in Doom textures. It's no fun to tweak shades of dark brown on a cyan background.

  To reuse images done with/for a previous version of DeuTex, you need to either invoke DeuTex with "-rgb 0 255 255" or convert your images by replacing all occurrences of colour (0, 255, 255) by colour (0, 47, 47). To preserve compatibility with WinTex, I didn't change the default transparent colour in WinTex mode ; it's still (0, 255, 255).

• Graphics: fixed segfaults due to bug in conversion of bitmap images to Doom pictures. Occured in certain 2-pixel high images such as STBFN045 in the Strife iwad.
• Graphics: now supports pictures and textures up to 509 high (was limited to 128).
• Graphics: now supports pictures and textures up to 1024 wide (was limited to 512).
• Graphics: new option "-pf" to deal with the different picture format in the Doom alpha iwad (the underlying functionality is not implemented yet !)
• Graphics: using -ppm does not cause anymore DeuTex to abort with "Bug: *** psvit ***".
• Graphics: fixed "-ppm" message.
• Sound, platform: a bug that must have prevented reading .wav files on big endian machines has been squashed.
• Command line: options can't start with a slash ("/") anymore. I don't think anyone used it and was a silly feature for a Unix program.
• Command line: not case insensitive anymore.
• Command line: changed the wording of error messages to use "option" instead of "command".
• Command line: added options "-?" and "--help" as synonyms for "-help".
• Doc: new option "-man" to print help in troff -man source format for inclusion in the man page.
• Doc: the version number is now a free-form string.
• Doc: Made on-line help more compact.
• Doc: updated "making.txt".
• Doc: made a proper README file.
• Makefile: renamed "unix" target as "strip".
• Makefile: new target "install".
• Makefile: new target "dist".
• Platform: reworked the handling of endianness. DeuTex used to deal with that through a set of macros that swapped bytes whenever the required endianness was not the same as the native endianness. To known the native endianness, DeuTex relied on a macro defined via -D.

  There were two problems with this scheme. Firstly, Olivier got the meaning of "little endian" and "big endian" backwards and defining LITTLE_ENDIAN in fact caused DeuTex to believe it was being compiled for a big endian machine. As the glibc headers happen to define LITTLE_ENDIAN if the machine is little endian, compiling DeuTex on a glibc little endian Linux system was impossible unless you made changes to the source.

  The other, more fundamental, objection against the old approach is that, as it needed the user to tell it about the native endianness by modifying the makefile, it prevented unattended builds and made things difficult for naive users.

  The new method eliminates this problem by using a different algorithm that does not need to know the native endianness. The end result is that you don't have to worry about endianness anymore.

• Internal: in TXTinit(), removed useless "% 0xFF" in index of TXTval.