IRC logs for #opendune on OFTC at 2010-03-08
⏴ go to previous day
00:01:06 <TrueBrain> when locking, I have to return a 32bit pointer
00:01:23 <TrueBrain> but I get that back at some point, assigned to handler '0', which means, conventional memory
00:08:56 <TrueBrain> it helps if I just disallow locking :p
00:09:21 <TrueBrain> but it seems to disable XMS then :p
00:21:11 <TrueBrain> it tries to be clever when it thinks the 32bit address is within the normal scope
00:21:35 <TrueBrain> and for some reason it reads the 32bit value as BE ...
00:22:03 <TrueBrain> that said, it still fails
00:22:28 <TrueBrain> which is funny, as it only copies bytes to the XMS so far
00:24:42 <TrueBrain> it does work outside the jit :)
00:26:41 <TrueBrain> I _think_ it screws with the overlay manager in yet another way :)
00:37:50 <TrueBrain> @calc 0x83f9244 - 0x8258820
00:37:56 <TrueBrain> @base 10 16 [calc 0x83f9244 - 0x8258820]
00:43:33 <TrueBrain> http://svn.opendune.org/~truebrain/libemu.patch <- glx: it doesn't work, but I can't figure out why not. I can't see anything wrong, yet it fails to work. It just randomly tries to return to 0000:0000 (which is in all cases WRONG). The odd thing is, it so far only wrote pieces of code to the EMS ...
00:44:15 <glx> oh you removed my JIT stuff ?
00:44:24 <TrueBrain> yes, it doesn't work when you are not resuming :)
00:44:29 <TrueBrain> so the solution is not 100%
00:44:39 <TrueBrain> I won't commit it, as we use the JIT for resuming more often
00:44:53 <TrueBrain> although with this EMS stuff that will be much more harder :)
00:45:19 <TrueBrain> (read: only for small missing cases it will work; otherwise XMS has to be disabled :p)
00:45:35 <TrueBrain> but I want to get the code for the voices decompiled
00:45:40 <TrueBrain> then we can disable XMS, doesn't matter
00:47:58 <TrueBrain> B4FC jumps to a new cs 8D16 ... sounds ... wrong
00:48:07 <TrueBrain> already fixed that seconds after posting
00:49:26 <TrueBrain> it is nothing near final
00:49:31 <TrueBrain> it doesn't have any propsets or what ever
00:49:35 <TrueBrain> I just want it to get it to work first :)
00:49:42 <TrueBrain> so ignore all the code, coding style, .. :)
00:49:57 <TrueBrain> B4FC handles the ini stuff ... it seems to fail there in a way it didn't before .. hmm ...
00:50:18 <TrueBrain> and then things start to break
00:51:59 <TrueBrain> XMS is known to be very scares in his handlers
00:52:08 <TrueBrain> so 2, at most 3 handlers per application is the general rule
00:52:22 <TrueBrain> (it is adviced to only use one, if possible)
00:53:21 <TrueBrain> again, please ignore the code
00:53:25 <TrueBrain> we can talk about that later if you like :)
00:53:29 <TrueBrain> for now it just has to work first :)
00:55:44 <TrueBrain> (and yes, calloc is C89)
01:01:06 <TrueBrain> then what kind of weird stuff does it pull ...
01:01:57 <TrueBrain> there, made the patch more commitable
01:02:01 <TrueBrain> just one printf debug statement
01:02:12 <TrueBrain> it seems, the 'emu_cs' gets corrupted
01:02:59 <glx> maybe not related to your changes
01:04:47 <TrueBrain> I don't touch the stack
01:04:51 <TrueBrain> I don't touch unneeded registers
01:05:00 <TrueBrain> yet ... it does something to screw things up
01:05:26 <glx> maybe one emu_ wrapper in opendune
01:05:36 <TrueBrain> I run completely in jit
01:05:41 <TrueBrain> so what ever we did, it doesn't matter
01:06:39 <glx> emu_get_memory8(emu_es, emu_bx, 4) = 0xCB; <-- maybe this then
01:06:53 <TrueBrain> I doubt it .. I make it a retf
01:06:58 <glx> the only thing that could play with stack
01:07:24 <TrueBrain> which should not happen
01:07:30 <TrueBrain> remember that the reti is also inserted by me
01:08:28 <TrueBrain> with a far jump, not really an other posibility
01:09:30 <glx> (because INT includes pushf ;) )
01:10:03 <TrueBrain> euh, okay .. the cs is valid
01:10:12 <TrueBrain> that is, it is returned by the overlay manager
01:10:24 <TrueBrain> and although the values are more unusual then we are used to
01:11:00 <glx> overlay used a different "page" ?
01:11:33 <TrueBrain> yeah, but well in range of what the JIT expects them to be
01:11:55 <TrueBrain> but what I don't get, is that the JIT in fact ignores all of it, even if it is wrong
01:11:57 <TrueBrain> it will just record wrong
01:14:20 <TrueBrain> Dune2 does some sanity checks on the replies from the XMS
01:14:45 <TrueBrain> k, enough for today ..
01:14:51 <TrueBrain> glx: can you see if MSVC accepts my patch?
01:14:56 <TrueBrain> that would be great
01:15:12 <TrueBrain> and if you can give it a spin in your JIT, and see if you can find anything odd/strange/weird/stupid/wrong/whatever :p
01:15:16 <TrueBrain> for now, good night
01:22:39 <glx> compile failed (new files not in the project ;) )
01:23:02 <glx> well libemu compiles, but linking to libemu.lib fails
09:18:10 <TrueBrain> glad glx told me how it failed :p
10:32:23 *** ChanServ sets mode: +v Xaroth|Work
11:24:55 <TrueBrain> hmm .. I think I know the XMS problem ...
11:25:31 <TrueBrain> JOY JOY HAPPY HAPPY JOY JOY HAPPY HAPPY JOY JOY
12:35:22 <DorpsGek> SVN: truebrain (r948) [JIT] -Add: a couple new functions (XMS support)
12:37:08 <DorpsGek> SVN: truebrain (r949) -Update (r948): update decompiled code
12:37:12 <TrueBrain> pff, almost r1000 ...
12:40:36 <TrueBrain> LOL! Now at start of OpenDUNE (with MT-32 and voices), I get: initializing MT-32 ... tick .. tick .. tick (takes a while)
12:44:51 <TrueBrain> bah, voices are disabled by configure
12:45:37 <TrueBrain> the MT32 has no voice support
12:45:44 <TrueBrain> for that .. we need other playback devices
12:45:47 <TrueBrain> something for glx to do :)
12:49:52 <TrueBrain> nowI will have to wait for glx :)
12:57:40 <Xaroth|Work> speaking of the devil!
13:02:07 <TrueBrain> I have some good nice: XMS works :)
13:02:16 <TrueBrain> just you said you had link problems? So I was afraid to commit :)
13:02:22 <TrueBrain> also, MT32 doesn't have a voice driver ..
13:02:51 <Xaroth|Work> some good nice? :o
13:03:04 <glx> link problems due to missing files in libemu project ;)
13:03:11 <glx> but commit and I'll update
13:03:15 <TrueBrain> you never made a generate script for it :p
13:03:39 <TrueBrain> and I have trouble starting the game with sound now
13:06:12 <TrueBrain> own, and crash.bins now became pretty useless
13:06:14 <TrueBrain> as the EMS is not in them
13:06:29 <TrueBrain> should be relative easy to add it, as it is a simple method
13:06:34 <TrueBrain> but .. not really important :p
13:07:17 <TrueBrain> mind proofreading it? :)
13:07:56 <TrueBrain> owh, duh, of course it fails
13:08:00 <TrueBrain> I forgot to revery a piece of code :p
13:08:25 <DorpsGek> SVN: truebrain (r950) -Fix (r949): committed one thing too many
13:08:47 <TrueBrain> you will have to compile in voice support yourself, as I don't know how you want to do that :)
13:08:50 <TrueBrain> which driver and stuff
13:11:12 <TrueBrain> did you btw know that when you select MT32 from the setup, you get driver 6?
13:11:21 <TrueBrain> which, with XMS, loads itself into the EMS (if I read this correct)
13:13:50 <TrueBrain> a tiny bit more detail?
13:13:53 <glx> `f__0070_07D0_0002_C33A' and `f__1FB5_017A_001C_0508'
13:15:51 <TrueBrain> 1FB5 .. wasn't that file stuff?
13:16:18 <glx> file access wrappers IIRC
13:16:37 <TrueBrain> hmm .. yes ... and a part I removed .. so .. this is kind of important to know what it does :p
13:17:01 <TrueBrain> of course I removed all my notes for it
13:19:05 <TrueBrain> function seems 'important'
13:19:17 <TrueBrain> from the opened PAK files it seems to find the biggest size
13:19:58 <TrueBrain> well .. I am not sure :p
13:20:01 <TrueBrain> it checks for a flag
13:20:06 <TrueBrain> and I hav eno clue anymore what that flag did
13:21:06 <TrueBrain> it is afunction with a uint32
13:21:15 <TrueBrain> and if some value of those files is bigger, it returns that value
13:21:18 <TrueBrain> else the parameter given
13:21:22 <TrueBrain> but it seems the first never happened
13:21:30 <TrueBrain> so .. lets make it an Obsolete statement :p
13:23:10 <DorpsGek> SVN: truebrain (r951) -Fix (r949): a few compile issues
13:23:44 <DorpsGek> SVN: truebrain (r952) -Fix (r951): compile before commit
13:25:22 <DorpsGek> SVN: truebrain (r953) -Fix (r951): clean compile before commit (dep-system breakage)
13:26:16 <DorpsGek> SVN: truebrain (r954) -Fix (r953): lazy ass do full compile before commit
13:26:51 <DorpsGek> SVN: truebrain (r955) -Fix (r951): OMG COMPLETE FAIL
13:26:52 <Xaroth|Work> going this rate reacing r1000 will be a breeeeze :P
13:27:03 <TrueBrain> 6 correction commits ....
13:27:12 <Xaroth|Work> and then you complain about my correction commits :P
13:27:25 <TrueBrain> this ... I never had before this bad
13:27:28 <TrueBrain> dunno what happened :(
13:27:29 <glx> when he fails he fails well ;)
13:27:37 <Xaroth|Work> yeh, you succeed at failing, TrueBrain :)
13:27:37 <TrueBrain> I blame sleep-shortage
13:28:07 <Xaroth|Work> grep is sooooo sloooow :o
13:28:22 <Xaroth|Work> i mean, it's only a grep -ri on around 40000 files!
13:28:44 <TrueBrain> you got to love IO cache :)
13:29:07 <glx> yeah like openttd startup time with many grfs
13:30:37 <DorpsGek> SVN: truebrain (r956) [JIT] -Add: support for music driver 6 (which is almost like 7)
13:30:46 <DorpsGek> SVN: truebrain (r957) -Update (r956): update decompiled code
13:30:51 <TrueBrain> I just had to do that :)
13:30:57 <TrueBrain> try 6, it is nice :p
13:32:03 <DorpsGek> SVN: truebrain (r958) [LibEMU] -Add: XMS support (on by default)
13:32:26 <glx> same as 7 but different ;)
13:32:32 <TrueBrain> glx: but I am afraid for voices you do need to add some other driver ... :(
13:32:41 <TrueBrain> glx: other extension, and minor difference in initalization
13:32:45 <TrueBrain> it waits for something ...
13:33:50 <TrueBrain> now XMS was much easier than expected :)
13:34:20 <glx> ho svn up did complain about xms.* existence
13:34:36 <glx> maybe because they were svn added ;)
13:34:36 <TrueBrain> that happens when you have a patch applied which has new files :)
13:37:39 <TrueBrain> 19 new functions .. not good for the stats :)
13:39:23 <glx> Avertissement,...1,...warning C4013: 'Handler_XMS' non défini(e) ; extern retournant int pris par défaut,...d:\developpement\opendune\trunk\decompiled\cs__217e.c,...2366
13:39:23 <glx> Avertissement,...2,...warning C4013: 'Handler_XMS' non défini(e) ; extern retournant int pris par défaut,...d:\developpement\opendune\trunk\decompiled\cs__2bf2.c,...77
13:41:19 <glx> don't tell me you can't understand the message
13:41:42 <glx> driver 6 is nice for music, but sounds are weird
13:41:55 <TrueBrain> something about 'extern'
13:43:07 <DorpsGek> SVN: truebrain (r959) -Fix (r955): still fail
13:46:15 <TrueBrain> but okay .. this should be sufficient to get you on the way glx ;) Maybe this way we can have voices for 0.3? :)
13:48:14 <glx> indeed music 6 is better than music 7
13:48:24 <glx> the problem is really sound 6
13:48:27 <TrueBrain> owh, I didnt check that :)
13:48:37 <TrueBrain> I just noticed the SETUP.EXE gives music 6 when selecting MT32
13:48:44 <TrueBrain> and I noticed it had a VERY small difference
13:48:57 <glx> the sand wind in intro is not good
13:49:51 <glx> but it's not possible to mix 6 for music and 7 for sound ;)
13:50:22 <glx> so it uses music for sound
13:50:57 <glx> ,...if (music->dfilename.csip != 0x0 && !strcasecmp((char *)emu_get_memorycsip(music->dfilename), filename)) {
13:50:57 <glx> ,...,...memcpy(sound, music, sizeof(Driver));
13:51:28 <TrueBrain> well, easy to 'fix' ;)
13:51:37 <glx> but not important for now
13:52:53 <glx> now which voice driver to choose
13:54:22 <TrueBrain> I have no clue which one would be easier
13:55:18 <TrueBrain> but without doubt one will be easier to implement than the other
13:55:28 <TrueBrain> like easier IN and OUT
13:57:13 <glx> 5 (ibmbak.com) doesn't exist either
13:59:54 <glx> I think I'll run JIT for all and keep the "simplest"
14:11:10 <glx> I guess all drivers will need that
14:13:19 <TrueBrain> BC, not listed here
14:13:28 <TrueBrain> B9 LAN RECEIVER.COM
14:13:29 <TrueBrain> BF LAN REDIRIFS.EXE
14:13:38 <TrueBrain> but either way, it is like the XMS
14:15:19 <glx> Media Vision Pro Audio Spectrum(TM) Digital Sound <-- that's 3
14:15:46 <TrueBrain> clearly needs some .sys loaded :)
14:19:47 <glx> because ibmbak.com clearly doesn't exists
14:19:48 <TrueBrain> application BC is reserved for network stuff :p
14:21:33 <TrueBrain> found some documentation on the Media driver ... lets not use it :p
14:22:50 <TrueBrain> 38A gives the suggestion it wants to do IRQ
14:22:55 <glx> hmm voices driver is in AB00 it seems
14:23:28 <glx> I fear some conflicts with music/sound
14:23:44 <TrueBrain> I was wondering a moment ago if we could decompile multiple music drivers
14:23:51 <TrueBrain> somehow TOC needs to know which one is being called :p
14:25:53 <glx> indeed voice overwrites music/sound
14:26:16 <TrueBrain> disable music/sound, move away AB00, and get voice to work
14:26:22 <TrueBrain> then move AB00 back :p
14:26:27 <TrueBrain> or rename it to AB01 :p
14:26:32 <TrueBrain> (and all AB00 entries :p)
14:57:17 <TrueBrain> eaxctly; as long as we don't have to compile in new functions, you will be fine :)
14:58:02 <TrueBrain> well, in theory you can even share the AB00, it just puts all sound and music functions in one file, without you know which one is which :p
15:00:18 <glx> but it may overwrite existing stuff
15:01:16 <TrueBrain> no, the JIT is build to detect those things
15:01:23 <TrueBrain> it can only get very confused when long blobs are equal
15:07:50 <glx> to ease work I could commit this
15:08:58 <glx> and do the same move in JIT so toc is happy ;)
15:09:15 <TrueBrain> don't forget to translate all the AB00 -> AB01 :p
15:09:19 <TrueBrain> there are plenty :p
15:10:05 <glx> but then jit will fail for unresolved music/sound stuff
15:48:59 <DorpsGek> SVN: glx (r960) -Change: move AB00 to AB01 to ease work on voice drivers
15:49:26 <DorpsGek> SVN: glx (r961) [JIT] -Change: move AB00 to AB01 to ease work on voice drivers
15:53:37 <TrueBrain> grep for AB00, and make sure they are all gone :)
15:53:42 <TrueBrain> list for AB00, and make sure they are all gone :)
16:00:19 <glx> better, now it doesn't readd cs__AB00.c, but removes cs__AB01.c
16:01:19 <TrueBrain> that might explain :p
16:01:22 <TrueBrain> did you add it with SVN? :)
16:04:44 <glx> ie 3 stupid (shouldn't happen) changes and a 44AF to AB01 revert
16:35:06 <glx> as I though, drivers 5 and 6 do nothing :)
16:35:33 <glx> except resolving some unresolved jumps
16:45:10 <DorpsGek> SVN: glx (r962) [JIT] -Add: a function (voice driver init)
16:45:58 <DorpsGek> SVN: glx (r963) -Update (r962): update decompiled code
16:46:25 <glx> that's all I get with drivers 5 and 6 (inexistant files ;) )
18:03:51 <glx> hmm in dosbox no voices with driver 4
18:08:35 *** DorpsGek` has joined #openDune
18:08:35 *** ChanServ sets mode: +o DorpsGek`
18:08:44 <glx> so the choice is SB or SBpro
18:12:20 *** DorpsGek` is now known as DorpsGek
18:44:43 <glx> of course toc doesn't resolve outside decompiled ;)
20:21:24 <glx> nice for now the code seems to match the doc I found
20:22:55 <glx> but of course the driver disable itself because libemu is just basic :)
20:24:04 <TrueBrain> that is how it is programmed :)
20:24:20 <TrueBrain> I love that about DOS
20:24:25 <TrueBrain> you can just return: no, I don't support that
20:24:29 <glx> I know I added some of them
20:24:31 <TrueBrain> and the app is like, well .. I pick this route then
20:26:57 <glx> and I can't find any doc about that
20:27:48 <glx> 226 seems to be control port
20:45:03 *** planetmaker has left #openDune
20:51:32 <glx> it says exactly what the code does
20:51:54 <glx> like "Read the write-buffer status port (2xC) until bit 7 is cleared "
20:52:13 <glx> of course never cleared ;)
20:53:34 *** planetmaker has joined #openDune
20:53:34 *** ChanServ sets mode: +v planetmaker
20:53:51 <planetmaker> must have been a cmd+w which was meant for an xcode window
20:54:07 <TrueBrain> well, you are missed, that is one positive thing I guess :)
20:55:06 <planetmaker> it does its job. Mostly
20:56:18 <glx> it's just a pain to get the right xcode version
20:56:32 <planetmaker> yeah... chatzilla at home doesn't give as good a channel overview as xchat...
20:56:38 <planetmaker> glx: hm... is it?
20:56:47 <planetmaker> they're all available for download afaik
20:56:55 <planetmaker> 2.5, 3.0(?) and 3.2
20:57:08 <planetmaker> anyway, if you need an sdk, just tell me ;-)
20:57:14 <glx> but finding something on apple site is well...
20:57:26 <planetmaker> yes, I know what you want to say ;-)
20:59:12 <planetmaker> the typical "I've seen it here last time... where is it now?" :-P
20:59:53 <glx> msdn is nice too, all the API is like minimal version 2000
21:01:45 <glx> Minimum supported client Windows 2000 Professional <-- for waveOutOpen() but it works on win9x too ;)
21:04:44 <planetmaker> he, that is nice then
21:05:02 <planetmaker> that's different for OSX it seems :-(
21:05:23 <glx> for OSX there's a total new API for each version
21:12:33 <glx> and of course incompatible with previous one ;)
21:15:17 <tneo> glx, i don't hear any sounds :-(
21:15:24 <tneo> client 0: 'System' [type=kernel]
21:15:24 <tneo> client 14: 'Midi Through' [type=kernel]
21:15:27 <tneo> client 16: 'TerraTec DMX6Fire' [type=kernel]
21:15:31 <tneo> 0 'MIDI-Front DMX6fire 0'
21:15:33 <tneo> 32 'Wavetable DMX6fire 0'
21:15:35 <tneo> client 128: 'OpenDUNE - Pre v0.3' [type=user]
21:15:39 <tneo> client 129: 'TiMidity' [type=user]
21:16:16 <TrueBrain> 2) <delete>, glx already said it :)
21:17:13 <TrueBrain> help glx, if you like :p
21:17:30 <tneo> let me do debugging seems I'm good at that ;-)
21:17:47 <glx> btw for music you can use 6 or 7
21:18:43 <tneo> cool I never played it with music
21:19:02 <tneo> that didn't work on the 386 back then, only had speech :-D
21:19:41 <glx> the opposite was more usual
21:20:19 <tneo> yeah, my friend didn't have speech at all, had something to do with high mem or so, can't recall anymore
21:21:10 <TrueBrain> I had a nice battle yesterday to make our layer allow high mem stuff :)
21:21:16 <TrueBrain> so now glx has a job to do :p
21:23:34 <glx> hmm it writes 0xD3 to DSP write port
21:23:45 <glx> but I can't see D3 in the doc
21:24:12 <TrueBrain> so ignore it for a sec :)
23:59:29 <DorpsGek> SVN: glx (r964) [LibEMU] -Add: some fake ports needed for voice driver
23:59:39 <DorpsGek> SVN: glx (r965) [JIT] -Add: started to map a voice driver
continue to next day ⏵