DOSBox DOS Emulator

Tracker: Bugs

5 Daggerfall fails to work with AMD64 Builds - ID: 1180880

Last Update: Comment added ( qbix79 )

Details:

With the AMD64 Native version of DOSBox, attempting to
run Daggerfall results in an Internal Error with an
error message stating that the program cannot access
data in either Dagger.snd or ARCH3D.bsa. This occurs
with both the Daggerfall Demos as well as the Full
Version and in both 64-bit 0.63 and CVS.

Debug output does not provide much useful information
though I haven't had time to use some of the more
advanced debugging options, I only glanced at the debug
output. The game will crash upon selecting a province
your character is from; a time normally a sound is played.

The same Daggerfall install runs fine on 32-bit
versions of DOSBox.

Submitted:

Brian Hill ( damicatz ) - 2005-04-11 16:47

Priority:

Status:

Closed

Resolution:

Fixed

Assigned:

Peter Veenstra

Category:

cpu-realmode

Group:

0.63

Visibility:

Public

Comments ( 28 )

Date: 2005-04-28 15:29 Sender: qbix79 Logged In: YES user_id=535630 Closing this report. I commited a while ago. Glad we resolved the matter
Date: 2005-04-20 21:47 Sender: qbix79 Logged In: YES user_id=535630 Well I think we then found the problem. It would be nice to compare the 2 logs. (so all debug information in and one log without the reinterpret_cast.) To see if we can actually determine how it went wrong (as the resulting filepointer was the same in both cases) success: dx=19C ax=ED14 pos=19CED14 <== I will commit the reinterpret cast to the cvs tommorow.
Date: 2005-04-20 21:07 Sender: damicatz Logged In: YES user_id=373610 I took out the intepret cast and that broke it again if that's what your asking. I didn't take out the other stuff. I made a mistake on the address. It's http://catclub.dnip.net/damicatz/debugworks.txt
Date: 2005-04-20 20:50 Sender: qbix79 Logged In: YES user_id=535630 Can't acces the File (404 error) Could you determine if the reinterpret_cast fixes it or that the other stuff fixes it ? (src/dos/drivelocal.cpp:428) reads in the patched version: int ret=fseek(fhandle,reinterpret_cast<Bit32s>(pos),seektype); and in the regular version int ret=fseek(fhandle,*pos,seektype); It would be nice to know if that makes the difference. Else it got fixed by something else.
Date: 2005-04-20 20:27 Sender: qbix79 Logged In: YES user_id=535630 Can't acces the File (404 error) Could you determine if the reinterpret_cast fixes it or that the other stuff fixes it ? (src/dos/drivelocal.cpp:428) reads in the patched version: int ret=fseek(fhandle,reinterpret_cast<Bit32s>(pos),seektype); and in the regular version int ret=fseek(fhandle,*pos,seektype); It would be nice to know if that makes the difference. Else it got fixed by something else.
Date: 2005-04-20 20:19 Sender: damicatz Logged In: YES user_id=373610 I applied the patch and was able to play the game with it. After taking it back off the game did not work again. Will test it with sound now. Log of working game : http://catclub.dnip.net/damicatz/logworks.txt
Date: 2005-04-20 19:52 Sender: qbix79 Logged In: YES user_id=535630 Yes the reading of 0 samples is indeed what is causing the problems. for reference: 27143532 Apr 14 11:12 ARCH3D.BSA My guess with that seek was that the seek was causing the read problems. Anyway an additional patch. Should be used together with the old one. Index: drive_local.cpp =================================================================== RCS file: /cvsroot/dosbox/dosbox/src/dos/drive_local.cpp,v retrieving revision 1.56 diff -u -r1.56 drive_local.cpp --- drive_local.cpp 25 Mar 2005 09:11:08 -0000 1.56 +++ drive_local.cpp 20 Apr 2005 19:48:26 -0000 @@ -392,9 +392,13 @@ //TODO Maybe use fflush, but that seemed to fuck up in visual c bool localFile::Read(Bit8u * data,Bit16u * size) { - if (last_action==WRITE) fseek(fhandle,ftell(fhandle),SEEK_SET); + if (last_action==WRITE) {fseek(fhandle,ftell(fhandle),SEEK_SET); LOG_MSG("la=write");} last_action=READ; size=fread(data,1,size,fhandle); + if((size)==0){ + LOG_MSG("error: %X",ferror(fhandle)); + LOG_MSG("eof: %X",feof(fhandle)); + } return true; }; @@ -420,8 +424,11 @@ //TODO Give some doserrorcode; return false;//ERROR } - int ret=fseek(fhandle,pos,seektype); + LOG_MSG("before %X",ftell(fhandle)); + int ret=fseek(fhandle,reinterpret_cast<Bit32s>(pos),seektype); + LOG_MSG("afterwards %X error %X" ,ftell(fhandle),ferror(fhandle)); if (ret!=0) { + LOG_MSG("UUUUUUUUUUUUUUUUUUUUUUUUUUUUSED!"); // Out of file range, pretend everythings ok // and move file pointer top end of file... ?! (Black Thorne) fseek(fhandle,0,SEEK_END); ---------------------------------------------------------------------- As you can see I put it in a small reinterpret cast. Feel free to remove it. but I think it might solve the problem. So if it works please test once without it. But if it still fails just attach the logfile in that case. fseek(fhandle,reinterpret_cast<Bit32s>(pos),seektype); used to be fseek(fhandle,*pos,seektype);
Date: 2005-04-19 21:50 Sender: damicatz Logged In: YES user_id=373610 Again I don't know much about DOS but perhaps this is odd as well : In various places my log will say : read in 936 out 0 And on yours it'll say read in 936 out 936. or some other value. But for some reason my version is outputting stuff to 0.
Date: 2005-04-19 21:18 Sender: qbix79 Logged In: YES user_id=535630 what struck my eye was that in the int1 picture the fseek is clearly present. seek cx=FFFX ax=XXXX pos=FFFXXXXX this is really tricky. I checked it out. dos wants the input variant to a 32 bit signed number and the output to be 32 unsigned number We do it both times 32 bit unsigned. the FFFFEXXXX is a negative number as far as I know so it's setting the filepointer back. (but maybe there the 32/64 bit stuff kicks in). but the resulting pos seems to fine so it's indeed a bit odd. anyway we know in which area to look. I will try to create a patch that logs reads (and the reason why it fails) extensively. I noticed that it fails continously if a seek with a high number (or negative number) is done (the reads afterwards). http://www.delorie.com/djgpp/doc/rbinter/id/10/28.html
Date: 2005-04-19 20:15 Sender: damicatz Logged In: YES user_id=373610 I did a diff of the two logs. Few areas of interest (some areas highlighted for contrast) http://catclub.dnip.net/damicatz/int1.jpg http://catclub.dnip.net/damicatz/int2.jpg http://catclub.dnip.net/damicatz/int3.jpg http://catclub.dnip.net/damicatz/int4.jpg http://catclub.dnip.net/damicatz/int5.jpg http://catclub.dnip.net/damicatz/int6.jpg http://catclub.dnip.net/damicatz/int7.jpg On your side it's taking stuff in and putting it out with the same values. With the 64-bit compile it seems to be outputting some stuff to "0" instead.
Date: 2005-04-19 20:00 Sender: damicatz Logged In: YES user_id=373610 Ok I re-compiled it again with debug support. That must have been turned off when I re-ran the configure script. http://catclub.dnip.net/damicatz/debug642.log As for the RealSetVec stuff I just added did this in dos.cpp around line 333: case 0x25: /* Set Interrupt Vector */ cout << "0x25 called" << endl; cout << reg_al << endl; cout << ds << endl; cout << reg_dx << endl; cout << RealMakeSeg(ds,reg_dx) << endl; RealSetVec(reg_al,RealMakeSeg(ds,reg_dx)); break;
Date: 2005-04-19 20:00 Sender: damicatz Logged In: YES user_id=373610 Ok I re-compiled it again with debug support. That must have been turned off when I re-ran the configure script. http://catclub.dnip.net/damicatz/debug642.log As for the RealSetVec stuff I just added did this in dos.cpp around line 333: case 0x25: /* Set Interrupt Vector */ cout << "0x25 called" << endl; cout << reg_al << endl; cout << ds << endl; cout << reg_dx << endl; cout << RealMakeSeg(ds,reg_dx) << endl; RealSetVec(reg_al,RealMakeSeg(ds,reg_dx)); break;
Date: 2005-04-19 08:10 Sender: qbix79 Logged In: YES user_id=535630 it would appear that you didn't compile dosbox with the debugger enabled. I really need it as It is very hard to compare the logfiles this way. You can leave your own debugging messages in it. But without the additional logging information provided by the debugger I don't have enough information. The sequence: 9519 0x25 called / 3 576 Reg_AL 3 ds 576 Reg_DX Hello Random Seperator 3355443776 RealSetVec Returns This ---------------------- is allready in the shutdown process you can see the shutdown process by this: int 21 3eXX in a loop (which is the closure of all open files. in my proper shutdown it reads: int 21 call 3E07 int 21 call 3E00 int 21 call 3E08 int 21 call 3E09 int 21 call 3E0A int 21 call 3E0B (as does it yours after the error. quoting from the first debug.txt ---------- 122872688: FILES:file open command 0 file c:\dagdemo2\arena2\map100i0.img int 21 call 4400 int 21 call 4201 int 21 call 4202 int 21 call 4200 int 21 call 3F41 int 21 call 3E0D int 21 call 3E07 int 21 call 3E00 int 21 call 3E08 int 21 call 3E09 int 21 call 3E0A int 21 call 3E0B -------------------- The 3E0D is normal. But then it decides to shutdown. so something above this is the problem. Hence I added information to int 21 call 4201 int 21 call 4202 int 21 call 4200 int 21 call 3F41 those calls. But without the debuggerlog I have a very hard time figuring out which file it used. The value is indeed a bit large. To make me understand it could you post the changes you made ? I will attach my debug log
Date: 2005-04-18 20:50 Sender: damicatz Logged In: YES user_id=373610 I added the patch, however for some reason it wasn't logging to the log file like the config file said it should so I did a redirected output again. http://catclub.dnip.net/damicatz/debug2.txt You'll have to ignore some of the random things about bob and such as I was doing some of my own testing but it shouldn't affect your patch as it was all using cout. RealSetVec is returning what appears to be an abnormally high value, I marked it in the log with messages surrounded by ** after the line in question. I'm not a DOS/Assembler coder so perhaps this is normal but 3355509121 seems large.
Date: 2005-04-18 20:20 Sender: qbix79 Logged In: YES user_id=535630 Well we have some small difference but I don't think it will matter. you seem to have (or not have) a arena2\berthesa.inf (or something like that..) I will attach my log as well. The last actions before your crash: 3 fileseeks and one read. ----A new patch------------ Index: dos.cpp =================================================================== RCS file: /cvsroot/dosbox/dosbox/src/dos/dos.cpp,v retrieving revision 1.83 diff -u -r1.83 dos.cpp --- dos.cpp 25 Mar 2005 12:06:48 -0000 1.83 +++ dos.cpp 18 Apr 2005 20:19:25 -0000 @@ -47,7 +47,7 @@ DOS_PSP psp(dos.psp()); psp.SetStack(RealMake(SegValue(ss),reg_sp-18)); } - + LOG_MSG("int 21 call %X",reg_ax); char name1[DOSNAMEBUF+1]; char name2[DOSNAMEBUF+1]; switch (reg_ah) { @@ -461,6 +461,7 @@ if (DOS_CloseFile(reg_bx)) { CALLBACK_SCF(false); } else { + LOG_MSG("close failed %X",dos.errorcode); reg_ax=dos.errorcode; CALLBACK_SCF(true); } @@ -473,6 +474,7 @@ MEM_BlockWrite(SegPhys(ds)+reg_dx,dos_copybuf,toread); reg_ax=toread; CALLBACK_SCF(false); + LOG_MSG("read in %X out %X",reg_cx,reg_ax); } else { reg_ax=dos.errorcode; CALLBACK_SCF(true); @@ -505,9 +507,11 @@ case 0x42: /* LSEEK Set current file position */ { Bit32u pos=(reg_cx<<16) + reg_dx; + LOG_MSG("seek cx=%X dx=%X pos=%X",reg_cx,reg_dx,pos); if (DOS_SeekFile(reg_bx,&pos,reg_al)) { reg_dx=(Bit16u)(pos >> 16); reg_ax=(Bit16u)(pos & 0xFFFF); + LOG_MSG("succes: dx=%X ax=%X pos=%X",reg_dx,reg_ax,pos); CALLBACK_SCF(false); } else { reg_ax=dos.errorcode; -----------The rest the same-----------------------
Date: 2005-04-15 20:22 Sender: damicatz Logged In: YES user_id=373610 Well for some reason Sourceforge doesn't want to let me upload it so you can get it at this URL : http://catclub.dnip.net/damicatz/debug64.log
Date: 2005-04-15 20:20 Sender: damicatz Logged In: YES user_id=373610 I can't select Black Mark as the game crashes as soon as you select a provience at the new character screen. I have attached the latest log file.
Date: 2005-04-15 14:49 Sender: qbix79 Logged In: YES user_id=535630 Allright. We need to synchronise our copies. Please select for sound and sound effects =>no sound restart dosbox start game click away demo screens new game black marsh male (always choose the fastest options). till you get ingame exit that game. in your dosbox.conf you need the following lines: [dosbox] memsize=32 [log] logfile=/tmp/debug64.log [cpu] cycles=5000 [dos] ems=false apply this patch on src/dos/dos.cpp Post the resulting logfile here. I made a log file using the 32 bits version of dosbox and the latest cvs myself. Index: dos.cpp =================================================================== RCS file: /cvsroot/dosbox/dosbox/src/dos/dos.cpp,v retrieving revision 1.83 diff -u -r1.83 dos.cpp --- dos.cpp 25 Mar 2005 12:06:48 -0000 1.83 +++ dos.cpp 15 Apr 2005 14:46:11 -0000 @@ -47,7 +47,7 @@ DOS_PSP psp(dos.psp()); psp.SetStack(RealMake(SegValue(ss),reg_sp-18)); } - + LOG_MSG("int 21 call %X",reg_ax); char name1[DOSNAMEBUF+1]; char name2[DOSNAMEBUF+1]; switch (reg_ah) {
Date: 2005-04-14 10:50 Sender: damicatz Logged In: YES user_id=373610 You can get the Daggerfall Demo at http://www.uesp.net/file.shtml?dagger/files/dfbetony.zip
Date: 2005-04-14 08:46 Sender: qbix79 Logged In: YES user_id=535630 I will try to find the Demo as well. as I run a 32 bit version of dosbox. There are some odd things in there I want to see wether this game is really this odd. (as it opens the same file many times. it makes little sense. (at least to me)
Date: 2005-04-13 22:04 Sender: damicatz Logged In: YES user_id=373610 For some reason dosbox wasn't recognizing the logfile= command. perhaps I put it in the wrong section. I redirected stdio to a file however with debug mode enabled so it should provide similar information. I do not currently have a 32-bit binary with debug mode enabled. If you could provide one I could run debug on it as well.
Date: 2005-04-13 20:35 Sender: qbix79 Logged In: YES user_id=535630 hmm I would need the full logfile inorder to see what might be wrong. if you use the debug version you can create a new configfile file. that one has extra entries in it. one of the entries is logfile= set it to a reasonable file. If you could do it for the 32 and 64 bits version. if something nice occurs I will provide you with some instructions to get huge amount of relevant log messages (by adding some log messages to dos core).
Date: 2005-04-13 20:27 Sender: damicatz Logged In: YES user_id=373610 I tried that and I'm still getting the same error.
Date: 2005-04-13 08:00 Sender: qbix79 Logged In: YES user_id=535630 in the file drive_local.cpp in src/dos around line 429 you notice an #if 0 try enabeling it and remove the line below the #endif -----now----- #if 0 fpos_t temppos; fgetpos(fhandle,&temppos); Bit32u * fake_pos=(Bit32u)&temppos; pos=fake_pos; #endif pos=(Bit32u)ftell(fhandle); ------------------- ----try--------- #if 1 fpos_t temppos; fgetpos(fhandle,&temppos); Bit32u * fake_pos=(Bit32u)&temppos; pos=fake_pos; #endif // pos=(Bit32u)ftell(fhandle); -------------
Date: 2005-04-12 21:15 Sender: damicatz Logged In: YES user_id=373610 How would I check to see if fseek is working?
Date: 2005-04-12 13:09 Sender: qbix79 Logged In: YES user_id=535630 hmm I would suspect an fileposition errror. you might want to check for fseek and friends to check if they behave correctly.
Date: 2005-04-12 01:41 Sender: damicatz Logged In: YES user_id=373610 I ran the debugger some more tonight and this is what it outputs right before the game crashes. 105397569: FILES:file open command 0 file c:\dagdemo\arena2\map100i0.img 105452384: MOUSE:Set sensitivity used with 49 49 105468357: PIC:0 mask D9 105468365: PIT:PIT 0 Timer at 90.00 Hz mode 3 105468369: PIC:0 mask D8 105468411: PIC:0 mask D9 105468649: PIC:0 mask D8 105469892: PIC:0 mask D9 105469900: PIT:PIT 0 Timer at 18.21 Hz mode 3 105469904: PIC:0 mask D8 105469945: PIC:0 mask D9 105470154: PIC:0 mask D8 105470247: PIC:0 mask F8 105470257: SBLASTER:DMA masked,stopping output, left 1357 105470327: PIC:0 mask F8 105472159: DOSMISC:DOS:Multiplex Unhandled call 1600 105472660: DOSMISC:DOS:Multiplex Unhandled call 1600 105482083: PIC:0 mask FC 106531424: MISC:MPU-401:Reset FF 110731845: PIC:0 mask F8 110732163: CPU:Illegal Unhandled Interrupt Called A 110759133: PIC:0 mask F9 110759162: PIC:0 mask F9 110759675: PIT:PIT 0 Timer set without new control word 110759675: PIT:PIT 0 Timer at 18.21 Hz mode 3 110759679: PIC:0 mask F8 110759708: PIC:0 mask F8 110773880: INT10:Set Video Mode 3 110773880: VGA:Blinking 8 111191646: FILES:file open command 0 file C:\DAGDEMO\ARENA2\BETHESDA.INF 111191900: DOSMISC:DOS:Multiplex Unhandled call 4A10 111191915: DOSMISC:DOS:Multiplex Unhandled call 4A10 111191930: DOSMISC:DOS:Multiplex Unhandled call 4A10 111191945: DOSMISC:DOS:Multiplex Unhandled call 4A10 111191960: DOSMISC:DOS:Multiplex Unhandled call 4A10 111191975: DOSMISC:DOS:Multiplex Unhandled call 4A10 111191990: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192005: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192020: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192035: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192050: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192065: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192080: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192095: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192110: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192125: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192140: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192155: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192170: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192185: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192200: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192215: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192230: DOSMISC:DOS:Multiplex Unhandled call 4A10 111192245: DOSMISC:DOS:Multiplex Unhandled call 4A10 111194241: VGA:H total 100, V Total 449 111194241: VGA:H D End 80, V D End 400 111194241: VGA:Width 640, Height 400, fps 70.086304 111194241: VGA:normal width, normal height aspect 1.000000
Date: 2005-04-11 16:49 Sender: damicatz Logged In: YES user_id=373610 By the way, I am running Gentoo Linux AMD64 2005.0 and the other games that I have tried (Descent, Dark Ages) work fine.

Attached Files ( 2 )

Filename	Description	Download
debug	Debug Log (64-bit)	Download
qbixdebug32.txt	my debug log with the changes in it.	Download

Changes ( 6 )

Field	Old Value	Date	By
status_id	Open	2005-04-28 15:29	qbix79
resolution_id	None	2005-04-28 15:29	qbix79
close_date	-	2005-04-28 15:29	qbix79
File Added	130863: qbixdebug32.txt	2005-04-19 08:10	qbix79
File Added	129750: debug	2005-04-13 22:04	damicatz
assigned_to	nobody	2005-04-12 13:09	qbix79

DOSBox DOS Emulator

Tracker: Bugs

Comments ( 28 )

Attached Files ( 2 )

Changes ( 6 )

About SourceForge

Find Software

Develop Software

Community

Help