IRC logs for #opendune on OFTC at 2011-08-15
⏴ go to previous day
04:13:40 *** Sonarpulse has joined #openDune
04:16:55 <Sonarpulse> a question: what is the best (english) version of dune to use with opendune
11:57:49 <DorpsGek> SVN: truebrain (r2367) -Codechange: small start removing the '>> 3' on X-coordinate based values
12:06:20 <DorpsGek> SVN: truebrain (r2368) -Codechange: gfx.c is now free of 'divided by 8' stuff, and all moved to higher layers
12:20:13 <DorpsGek> SVN: truebrain (r2369) -Fix: in case of reading a text file, make sure the string is null terminated
12:20:47 <DorpsGek> SVN: truebrain (r2370) -Codechange: the scenario INI file now no longer ends up in a global read buffer
12:24:46 <TrueBrain> hmm .. I just for lolz checked valgrind what is not released at quit
12:24:55 <TrueBrain> SDL seems to fail to unload a lot :P
12:25:55 <glx> and I properly quit SDL_Audio too
12:26:11 <TrueBrain> yeah, asound reports it is still active too :p
12:26:45 <glx> SDL_QuitSubSystem(SDL_INIT_AUDIO);
12:28:17 <TrueBrain> the trace of asound doesn't end in OpenDUNE, lolz
12:28:41 <TrueBrain> ah, comes via SDL_AudioInit
12:30:14 <TrueBrain> and clearly my valgrind is not up-to-date with dl :p
12:30:19 <TrueBrain> it reports errors there, which is normal :)
12:30:27 <TrueBrain> (valgrind has an ignore list for all libdls, but clearly not the one I am using)
12:32:18 <TrueBrain> meh, this makes it hard to find problems
12:32:31 <TrueBrain> well, problems .. to make us free nicely :)
12:33:29 <DorpsGek> SVN: truebrain (r2371) -Fix: free all timer related pointers on closing
12:38:15 <TrueBrain> weird stuff ... snd_seq_open / snd_seq_close also leaks :p
12:38:50 <TrueBrain> you didn't write asound? :p
12:39:44 <glx> does it alloc _midi even when it fails ?
12:40:12 <TrueBrain> don't think so; it also just calls closing it, so ... meh :P
12:40:29 <TrueBrain> and _midi is set there
12:40:33 <TrueBrain> so that is not the issue, sadly ..
12:41:54 <glx> it's alloca, but who knows
12:42:37 <TrueBrain> if I only do snd_seq_open() and snd_seq_close(), it leaks
12:44:44 <TrueBrain> searching for snd_seq_close on google puts OpenDUNE as 8th hit
12:44:49 <TrueBrain> that is a bit .. disturbing :P
12:46:09 <glx> hmm snd_seq_open broadcasts an event
12:52:03 <TrueBrain> opening and close it again, doesn't seem to change it .. so static buffers most likely
12:52:12 <TrueBrain> 39681 / 298 .. lets try it I am right ..
12:52:42 <TrueBrain> indeed ... opening and close 100 times doesn't change a thing
12:55:23 <planetmaker> a happy TrueBrain :-)
12:56:05 <TrueBrain> I went to all header files
12:56:10 <TrueBrain> found the functions that required no parameters
12:56:15 <TrueBrain> excluded 'sizeof' functions
12:56:21 <TrueBrain> 2 remaining, one called:
12:56:28 <TrueBrain> snd_config_update_free_global()
12:58:39 <DorpsGek> SVN: truebrain (r2372) -Fix: when closing down ALSA, it keeps global variables active. Free them too.
12:59:43 <TrueBrain> SDL has the same issue for me :p
13:01:57 <TrueBrain> if I manually add it (tricky) it does remove a lot of stuff .. but that is something that can't be committed :(
13:04:14 <TrueBrain> guess this is why I hate SDL etc :P
13:13:42 <glx> oh looks like I already downloaded a previous version
13:16:36 <DorpsGek> SVN: truebrain (r2373) -Fix (r2372): SDL also uses asound, and also does not free the globals, possibly causing corruption
13:22:58 <glx> compiles and link with vld, that's a good start :)
13:28:27 <DorpsGek> SVN: truebrain (r2374) -Add: a suppression file for valgrind to ignore SDL and ALSA 'leaks', something we will never be able to solve from our software (it is an downstream issue)
13:28:41 <TrueBrain> let of things to do in OpenDUNE itself :)
13:31:51 <DorpsGek> SVN: truebrain (r2375) -Fix: plug the biggest non-free'd-memory-on-close: the GFX screens
13:38:22 <DorpsGek> SVN: truebrain (r2376) -Fix: go nice on font mallocs and pointers
13:40:34 <glx> assert(voiceID < 120); <-- triggered when playing end game anim
13:41:36 <TrueBrain> <= maybe? Trace it :D
13:43:17 <glx> let's make it an int8 instead of int16 :)
13:49:01 <DorpsGek> SVN: glx (r2377) -Fix: replace an assert with a return
13:49:14 <DorpsGek> SVN: truebrain (r2378) -Fix: fix all remaining un-free'd pointers. Everything should be free'd on close now, so we can track any real leaks now (which are not present atm :P)
13:49:28 <TrueBrain> valgrind now returns an empty report :D
13:49:47 <glx> vld doesn't report anything, even when I deliberately add a leak
13:50:12 <TrueBrain> sounds like a horrible detection tool :D
13:51:44 <TrueBrain> hmm, looking back, it is not a real surprise we have no leaks yet ... we converted everything from a closed 16bit world, so the only leaks possibles is where we put mallocs, and we were well aware where we did that :D
14:02:35 <DorpsGek> SVN: truebrain (r2379) -Codechange: make sure all free()'d pointers are set to NULL to avoid weird problems later on in the development
14:05:30 <glx> if (memcmp(GFX_Screen_Get_ByIndex(0), s_gfx_screen8, SCREEN_WIDTH * SCREEN_HEIGHT) == 0) { <-- segfault
14:05:43 <TrueBrain> need a bit more info :D
14:06:01 <glx> GFX_Screen_Get_ByIndex returns NULL
14:06:17 <TrueBrain> there are 3 ways to exit :P
14:07:44 <TrueBrain> which file is that line in?
14:08:24 <DorpsGek> SVN: truebrain (r2380) -Fix: make sure the timer is removed before we remove the video, as the timer might still read video buffer
14:08:30 <TrueBrain> then I would guess that might fix it
14:10:29 <DorpsGek> SVN: truebrain (r2381) -Fix (r2380): oops, Video_Uninit does not free the videobuffer, GFX_Uninit does
14:13:19 <TrueBrain> what to do now .. euuhh ... :P
14:16:08 <TrueBrain> SmatZ: how did you detect which includes are still needed for OpenTTD? :)
15:04:46 <glx> [SCRIPT] [ERROR] Stack Overflow at d:\developpement\opendune\trunk\src\script\scrÌÌÌÌÌÌÌÌx>ÌÌÌÌ‚wÐõA; Type: Unit; Index: 51; Type: 3;
15:05:02 <glx> something looks weird in this line :)
15:06:35 <glx> hmm not surprising I think
15:07:27 <glx> Script_Error() is just plain wrong I think
15:08:06 <glx> at least buffer is too small ;)
15:10:00 <TrueBrain> bit short now yes :p
15:14:02 <DorpsGek> SVN: truebrain (r2382) -Codechange: bit weird to define g_scenario in security.c ..
15:19:34 <glx> "Visual Leak Detector. Seems to correctly find all leaks, but in my case it doesn't work since I have a 3rd party DLL that simply aborts the process at its DllUnload (seems to be a Windows 7 specific problem)." in some comment on stackoverflow.com
15:22:03 <glx> looks like I have the same problem
15:22:04 <TrueBrain> did you fix Script_Error btw?
15:22:43 <TrueBrain> please do if you have the time :D Buffer overflows are bad to have :P
15:24:04 <glx> vsnprintf should check it but it doesn't ;)
15:24:32 <glx> oh it does, but doesn't add the \0
15:24:36 <TrueBrain> maybe it does but doesn't add \0
15:24:43 <TrueBrain> which is _really_ annoying :p
15:25:04 <glx> backward compatibility ;)
15:25:12 <glx> there's a "secure" version
15:25:13 <TrueBrain> I believe it is _vsnprint or something?
15:25:18 <TrueBrain> maybe you can add it to os/ stuff?
15:37:24 <glx> snprintf has the same problem IIRC
15:37:30 <glx> let steal openttd code :)
16:04:51 <DorpsGek> SVN: glx (r2383) -Codechange: implement safer [v]snprintf() for MSVC
16:07:00 <glx> any objection is I increase buffer size ?
16:07:20 <TrueBrain> of course not, please do :)
16:07:27 <TrueBrain> it is way to small :)
16:08:56 <DorpsGek> SVN: glx (r2384) -Fix (r2350): buffer in Script_Error() was way too small
17:23:25 *** Sonarpulse has joined #openDune
17:35:29 *** planetmaker has joined #openDune
17:35:29 *** ChanServ sets mode: +v planetmaker
17:45:46 <DorpsGek> SVN: truebrain (r2385) -Fix (#83): the 'fp' could remain set to a non-NULL value when the file was in fact closed
17:48:36 <TrueBrain> happens in Dune2 too :p
17:48:43 <TrueBrain> as they work on the same assumption as we do: it cannot fail
17:48:48 <TrueBrain> so I doubt it fixes the problem
17:49:44 <TrueBrain> I just can't see how fclose can fail like that, but oaky
17:49:56 <TrueBrain> if r2385 really fixes the problem, the user has more problems :D
18:26:29 *** ChanServ sets mode: +v Yexo
18:58:44 <Sonarpulse> this is probably the most baddass project i have seen in a while
18:59:23 <Sonarpulse> does it often not work? does the config dos program help or hurt?
19:00:15 <glx> no need to run config dos program
19:36:09 <Sonarpulse> ok. Is no audio a common problem?
19:40:01 <Sonarpulse> then again i've never played Dune II so if the intro is silent I wouldn't know it.
19:40:26 <glx> there is music and voices in intro
19:40:59 <Sonarpulse> I tried the original for a second, and no audio there either
19:41:18 <Sonarpulse> probably just a silent version of the game
19:43:07 <Sonarpulse> put_dune2_here.txt
19:43:12 <TrueBrain> use a pastebin for that
19:44:01 <glx> I remember TrueBrain had that "problem"
19:44:35 <Sonarpulse> cool site (pastebin)
19:44:52 <glx> anything in error.log or output.log ?
19:44:54 <TrueBrain> useful for longer text over IRC. Keeps chats readable ;)
19:45:56 <Sonarpulse> should have figured, but my res is pretty high so didn't look as smammy
19:47:19 <TrueBrain> but yeah, if you use 0.6 with the Dune2 1.07EU datafiles, you should have sound on Windows
19:47:26 <TrueBrain> which windows version?
19:47:51 <Sonarpulse> same thing with file sizes
19:48:17 <TrueBrain> did you overwrite dune.cfg while copying?
19:48:20 <Sonarpulse> all *.logs are empty
19:48:35 <Sonarpulse> I did at first, but then I reverted it
19:48:39 <TrueBrain> make sure you use the dune.cfg supplied with the zip ..
19:50:09 <glx> opendune ignores dune.cfg for audio ;)
19:51:02 <glx> only language is still read from it
19:51:23 <TrueBrain> glx: but the ingame option is stored there, not?
19:51:37 <TrueBrain> or is that options.cfg?
19:52:01 <TrueBrain> so yeah, music should work ;)
19:52:14 <TrueBrain> Sonarpulse: during the intro, do you see subtitles when the emporer is talking?
19:52:15 <Sonarpulse> i will look for the other config file
19:52:29 <TrueBrain> so as far as our software cares, sound is working
19:52:37 <TrueBrain> make sure your default sound drive etc is set right?
19:53:05 <Sonarpulse> i get windows sounds
19:53:12 <Sonarpulse> drivers are sys default i think
19:53:51 <TrueBrain> I am out of ideas :)
19:54:25 <Sonarpulse> do you guys no of a simple SDL program i could download to see if it's that ends working?
19:54:25 <glx> I don't remember if vista has a sndvol.exe as nice as win7
19:55:06 <Sonarpulse> vista does to per program volume (probably just a dirty hack like most new vista features)
19:55:14 <Sonarpulse> and opendune is up
19:56:24 <glx> you should see sound level then
19:56:59 <Sonarpulse> hm if i click the slider on windows sounds or the master, it makes a beeb
20:03:26 <Sonarpulse> but not if I click on opendune
20:06:19 <glx> only master volume "beeps" on change
20:08:35 <Sonarpulse> btw does opendune still need duneII.exe now that the game is 100% decompiled
20:10:12 <TrueBrain> OpenDUNE only depends on the .PAK files of Dune2 1.07
20:11:52 <Sonarpulse> i think some readme in the zip still says the exe is required
20:12:00 <TrueBrain> it says 1.07 data files :)
20:12:08 <TrueBrain> leaves it a bit in the middle what exactly is required
20:12:55 <Sonarpulse> nah i just checked:
20:12:56 <Sonarpulse> Installation & Running
20:12:56 <Sonarpulse> ----------------------
20:12:57 <Sonarpulse> Copy the original Dune2 1.07 data files (including dune2.exe) to data/. All
20:12:57 <Sonarpulse> filenames should be lowercase for case-sensitive file systems (Linux, MacOSX).
20:12:59 <Sonarpulse> All three existing versions of the Dune 1.07 data files (eu, hs and us) will
20:12:59 <Sonarpulse> work, but only with the eu/hs data files the French language will work, and
20:13:01 <Sonarpulse> only with the eu data files the German language will work.
20:13:11 <TrueBrain> again, pastebin for those pastes please :)
20:13:28 <TrueBrain> everything over 3 lines, use pastebin :)
20:13:49 <glx> we say data files, we don't name them ;)
20:14:06 <TrueBrain> "including dune2.exe"
20:14:13 <TrueBrain> no clue why that didn't get removed; something for the next version ;)
20:14:30 <Sonarpulse> oh yeah. sorry, takes a while to build a new habit
20:15:06 <Sonarpulse> glad to help "debugging" the readme
20:15:48 <Sonarpulse> Do you guys plan on putting in multiplayer done the road?
20:16:06 <TrueBrain> maybe; but the game is heavily unbalanced for multiplayer
20:16:14 <TrueBrain> the game is not fair
20:16:26 <planetmaker> why wouldn't it, if both start the same?
20:17:05 <TrueBrain> because one house is stronger than the other :)
20:17:14 <TrueBrain> there is a heavy inbalance there, which in singleplayer is not really important :)
20:17:21 <glx> and the game is very random :)
20:17:38 <TrueBrain> so multiplayer has no priority of any kind :)
20:17:39 <glx> load the same game twice, different things will happen
20:17:41 <planetmaker> well, random is not necessarily bad; the battles are predicatable enough
20:17:50 <TrueBrain> glx: those things are fixable :)
20:18:01 <planetmaker> just the impact of the heavy rockets sucks ;-)
20:18:13 <Sonarpulse> litteraly relying on random numbers?
20:18:18 * planetmaker lost a few missions due to that
20:18:44 <Sonarpulse> or random lags and things throwing the game off?
20:18:58 <TrueBrain> when harkonnen ahs house missile ready
20:19:08 <TrueBrain> repeat till success
20:19:20 <TrueBrain> other way also works: you got shot by house missile of harkonnen?
20:22:07 <Sonarpulse> not like Tiberian dawn
21:33:31 <glx> oh no, we will need to fix it now :)
21:34:57 <TrueBrain> told him to make a bug report out of it so we would hav eto :P
21:35:40 <glx> and it's not upper case is not allowed
21:41:51 <glx> as always with MS, it's almost impossible to exploit :)
21:42:21 <glx> and vld still refuses to work
21:47:45 <TrueBrain> so valgrind and vld are in disagreement
21:48:19 <glx> but I dump in PrepareEnd()
21:48:40 <glx> while it starts to log before main()
21:48:56 <glx> let's try to tweak that a little
21:54:05 <glx> looks more like MSVC leak report
22:00:18 <glx> the leaks in Sound_Unknown0823() seems valid at first look
22:04:05 <TrueBrain> I don't have sounds, so that is possible :
22:04:13 <glx> in Driver_LoadFile() it's the driver->filename malloc
22:06:09 <glx> oh of course Sound_Unknown0823() is valid
22:07:18 <glx> g_variable_3E54[] items are freed before loading only
22:08:33 <glx> or there's a bug in the function :)
22:19:12 <glx> hmm no it's really a leak
22:42:51 <DorpsGek> SVN: glx (r2386) -Fix: pre-loaded voice files were not free'd on close
22:58:52 *** pm is now known as planetmaker
23:07:03 <DorpsGek> SVN: glx (r2387) -Fix: active sound/music was not free'd on close
23:29:30 <DorpsGek> SVN: glx (r2388) -Fix: DSP_Callback() could run with uninitialised data
23:54:10 <TrueBrain> glx then fix the crashdumps :D
23:54:44 <glx> anyway I fixed a random segfault at exit (at least for windows)
23:55:07 <TrueBrain> but if windows crashes, it doesn't leave anything by default?
23:57:00 <TrueBrain> well, I am off to bed; nn :)
continue to next day ⏵