[LeoD]

Crash Report - Short Version

On WinXP eduke32.exe crashes immediately after using a security monitor when
compiled with gcc 4.4.0-TDM-1-mingw32.

gcc 4.3.3-tdm-1-mingw32 is OK. Could someone please confirm/contradict this?


Crash Report - Long Version

Despite my lack of debugging knowledge I have tried to isolate it:

duke3d.cfg/eduke32.cfg settings do not affect this problem.
It occurs on both of my PCs : Pentium/Matrox, Athlon/NVIDIA.
It occurs with the latest source code (svn1461) as well as with the
oldest I managed to build (svn0016 from 2006).

After some makefile twiddling I found out that it does not happen when
setting OPTLEVEL from 2 to 0. OPTLEVEL?=1 lets the crash reappear.

After some object file juggling it seems to be related to engine.o.
To get a crash-free executable I compiled normally with gcc-4.4.0, then
replaced engine.o with a previously built version (OPTLEVEL?=0) and
re-linked the exe. Seems like a compiler bug to me.

I hope this information is helpful in some way.

[TerminX]

Well, there's an official MinGW GCC 4.4.0 now (as in not a build from tdragon.net) which you could try.

[supergoofy]

gcc 4.4.0 is causing problems

even if you compile successfully the source, the compiled binary causes crashes.


BETTER STAY WITH GCC 4.3.3

[supergoofy]

BETTER STAY WITH GCC 4.3.3

MinGW_gcc-4.3.3-tdm-1.rar (updated/repacked)
http://www.mediafire...php?deydv2znmzz

MSYS-1.0.11-rc-1.exe
http://www.mediafire...php?tihqwoowkzo


Please use gcc 4.3.3 to compile eduke32

[LeoD]

TX, on Jul 15 2009, 10:05 PM, said:

Well, there's an official MinGW GCC 4.4.0 now (as in not a build from tdragon.net) which you could try.

Same thing. MinGW users will indeed have to stick to 4.3.3 for now.

[supergoofy]

MinGW_gcc-4.3.3-tdm-1_libogg-1.1.4_libvorbis-1.2.3.rar

http://www.mediafire...php?gcnz3mmzzz4

[LeoD]

GCC 4.4.1 TDM-1 has been released - still not suitable for building EDuke.

[supergoofy]

I haven't tested it yet, because I don't have the needed free time.

[TerminX]

LeoD, on Aug 23 2009, 06:42 AM, said:

GCC 4.4.1 TDM-1 has been released - still not suitable for building EDuke.

What breaks?

[LeoD]

See my post from 2009-07-15 on top of this page.
Start E1L1, go to the nearest security monitor, use it and step back - crash.
Replacing eobj_win/engine.o with a 4.3.3 compiled version fixes that.

supergoofy, on Aug 23 2009, 06:43 PM, said:

I haven't tested it yet, because I don't have the needed free time.

Hey, it's Sunday!

[Todd]

LeoD, on Aug 23 2009, 05:05 PM, said:

See my post from 2009-07-15 on top of this page.
Start E1L1, go to the nearest security monitor, use it and step back - crash.
Replacing eobj_win/engine.o with a 4.3.3 compiled version fixes that.


Hey, it's Sunday!

My copy crashes on security monitors too--and I thought I had compiled with 4.3.3, but maybe I didn't.

You can use the monitor just fine--it's just that when you are done, instead of returning to DukeLand, it goes
to Cygwin Prompt land or wherever.

[TerminX]

Interesting. 4.4.x is definitely miscompiling something on win32 with optimizations enabled... a good place to start looking for culprits would be the code in engine.c used for drawing a frame to a tile with the software renderer.

[Stabs]

so whats the word on r1490?

anything new happening

[TerminX]

Committed some stuff today. Nothing major... mostly just minor feature requests and bug fixes that happened over the past couple weeks when I was without a proper monitor on my development box. I was able to get a broken Dell 2001FP off eBay for $30 (plus $17 shipping) however that while it ended up having a defective panel, I was able to MacGyver a fix for the problem out of paper scraps and pencil erasers (literally). So, while I don't expect this fix to hold for any extended length of time it does let my gamble on hardware in unknown condition pay off for a while until I can afford a monitor that won't flip out in a week or so.

[Danukem]

IIRC there was at least one bug fix that I consider major, so thanks for posting it!

[The Commander]

Generaly I can build the source fine, but as of late ive been getting these in the log.

include/polymer.h:319: warning: 'polymer_editorselect' declared 'static' but never defined
Built object ../eobj_win/polymer.o 
Built object ../eobj_win/mmulti_unstable.o 
Built object ../eobj_win/winlayer.o 
Changing dir to /c/eduke 
Changing dir to /c/eduke/source/jaudiolib 
Built object obj_win/drivers.o 
src/fx_man.c: In function 'FX_StartRecording':
src/fx_man.c:941: warning: unused parameter 'MixRate'
src/fx_man.c:942: warning: unused parameter 'function'
Built object obj_win/fx_man.o 
src/multivoc.c: In function 'RestoreInterrupts':
src/multivoc.c:172: warning: unused parameter 'a'
src/multivoc.c:281:5: warning: multi-line comment
src/multivoc.c: In function 'MV_StartRecording':
src/multivoc.c:2011: warning: unused parameter 'MixRate'
src/multivoc.c:2012: warning: unused parameter 'function'
src/multivoc.c: In function 'MV_PlayLoopedWAV':
src/multivoc.c:2284: warning: unused parameter 'ptrlength'
src/multivoc.c:2286: warning: unused parameter 'loopend'
src/multivoc.c: In function 'MV_PlayLoopedVOC':
src/multivoc.c:2499: warning: unused parameter 'ptrlength'
Built object obj_win/multivoc.o 
Built object obj_win/mix.o 
Built object obj_win/mixst.o 
Built object obj_win/pitch.o 
src/vorbis.c: In function 'close_vorbis':
src/vorbis.c:110: warning: unused parameter 'datasource'
src/vorbis.c: In function 'MV_PlayLoopedVorbis':
src/vorbis.c:318: warning: unused parameter 'loopend'
src/vorbis.c: At top level:
third-party/mingw32/include/vorbis/vorbisfile.h:75: warning: 'OV_CALLBACKS_DEFAULT' defined but not used
third-party/mingw32/include/vorbis/vorbisfile.h:82: warning: 'OV_CALLBACKS_NOCLOSE' defined but not used
third-party/mingw32/include/vorbis/vorbisfile.h:89: warning: 'OV_CALLBACKS_STREAMONLY' defined but not used
third-party/mingw32/include/vorbis/vorbisfile.h:96: warning: 'OV_CALLBACKS_STREAMONLY_NOCLOSE' defined but not used
Built object obj_win/vorbis.o 
src/driver_nosound.c: In function 'NoSoundDrv_ErrorString':
src/driver_nosound.c:32: warning: unused parameter 'ErrorNumber'
src/driver_nosound.c: In function 'NoSoundDrv_PCM_Init':
src/driver_nosound.c:37: warning: unused parameter 'mixrate'
src/driver_nosound.c:37: warning: unused parameter 'numchannels'
src/driver_nosound.c:37: warning: unused parameter 'samplebits'
src/driver_nosound.c:37: warning: unused parameter 'initdata'
src/driver_nosound.c: In function 'NoSoundDrv_PCM_BeginPlayback':
src/driver_nosound.c:46: warning: unused parameter 'BufferStart'
src/driver_nosound.c:46: warning: unused parameter 'BufferSize'
src/driver_nosound.c:47: warning: unused parameter 'NumDivisions'
src/driver_nosound.c:47: warning: unused parameter 'CallBackFunc'
Built object obj_win/driver_nosound.o 
src/driver_directsound.c: In function 'fillDataThread':
src/driver_directsound.c:150: warning: unused parameter 'lpParameter'
Built object obj_win/driver_directsound.o 
Changing dir to /c/eduke 
Linked executable eduke32.exe 
Built object obj_win/astub.o 
Built object obj_win/buildres.o 
Built object obj_win/sounds_mapster32.o 
Changing dir to /c/eduke/build 
Built object ../eobj_win/build.o 
Built object ../eobj_win/config.o 
Built object ../eobj_win/startwin.editor.o
Changing dir to /c/eduke 
Linked executable mapster32.exe 
Build successful:

Are these normal or is something out of date?

[supergoofy]

defined parameters that are not used. no harm at all. if you want delete them or put them inside comments ( /* */ )

[The Commander]

Everything works fine apart from that popping noise when using the highest sound bitrate. Ive seen a few other people report this as well.

[Skulldog]

Is everyone still using gcc 4.3.3 to compile eduke32's lastest svn?

[TerminX]

I am. 4.4.x still miscompiles part of the engine on Windows as far as I can tell, causing crashes when drawing the screen to a tile such as is done with savegames and security viewscreens.

[LeoD]

Did you take a look at what dungeonduke found out?
http://forums.duke4....?showtopic=1594
(My 4.4.1 exe still has problems, especially when loading maps or saved games.)

[bioman]

Hello,

Since r1586, building failed on Linux / gcc 4.4.x :

src/engine.c: In function ‘equation’:
src/engine.c:5893: internal compiler error: in tree_nrv, at tree-nrv.c:143
Please submit a full bug report,
with preprocessed source if appropriate.
See <file:///usr/share/doc/gcc-4.4/README.Bugs> for instructions.
Failed building ../eobj/engine.o from src/engine.c!
make[1]: *** [../eobj/engine.o] Erreur 1
make: *** [enginelib] Erreur 2

Seem to be related with this version of gcc only, it works with 4.3 :
make CC=gcc-4.3

[TerminX]

Yes, I also found another section of code last night that gcc is unfortunately miscompiling. It's best to stick with 4.3 at the moment. It looks like gcc 4.4.3 came out today, so I guess we'll see if that improves anything as well.

[NY00123]

I confirm that I've got the very same error as bioman has. Well excluding the fact that it says "Error" instead of "Erreur" and the like

If it matters, I'm using Ubuntu 9.10 for x86_64. Errors with default gcc, now I see it compiles with CC=gcc-4.3 as bioman has told.

[bioman]

It seems specific to 64-bit platforms it build fine on 32-bit.

[supergoofy]

Compiling environment for EDuke32:

MinGW_gcc-4.4.1-tdm-2.rar
http://www.mediafire...php?zm2mbzdqwkm

MinGW_gcc-4.5.0-1.rar
http://www.mediafire...php?4wjlhmjmh34

MSYS-1.0.11.exe
http://www.mediafire...php?qydxuzkytii

msysCORE-1.0.14-1-update.rar
http://www.mediafire...php?l3m2xqj522y


Note: If you use gcc-4.5.0-1 then you should follow the tip:
Just find and remove the "-ftlo" parameter in Makefile and build/Makefile of EDuke32's source.
In other words edit both "Makefile" files (including the one inside "build" folder), then search and
change the line:
debug += -flto
to the line:
# debug += -flto

Note: The gcc-4.4.1-tdm-2 doesn't need this tip, it compiles fine without any change in Makefile.


NEWS: From what I see here: http://sourceforge.n.../tdm-gcc/files/
We will soon have TDM's GCC/MinGW32 Builds for GCC 4.4.4

[TerminX]

No, 4.4.x doesn't need that change because it doesn't support LTO and therefore the makefile doesn't enable it.

[supergoofy]

Thanks TX for the info. I haven't checked the log for gcc 4.4.x, I will correct my above post.

[MusicallyInspired]

I screwed up my compiling environment for EDuke32 so I tried to make another one from these links. It compiles successfully but strangely much too fast. When I go to run either eduke32.exe or mapster32.exe I get errors saying it can't find procedures from the libssd library.

So I followed the directions on the Wiki instead and that makes MSYS unusable because it keeps looking for a msys-intl-8.dll file when I try to enter any command at all.

[MusicallyInspired]

Ok, so thanks then.