Re: wine-0.9.37 ktrace
I just realized you're right, most of the trouble are due to the errors you point out. (I certainly have a made a fool out of myself by now) However, when I run make depend, comment out the building of dnsapi in dlls/Makefile and then run make, it seems to complete and ends up with the message Wine build complete as well as building the wine-pthread binary. Michael Small wrote: On Sun, May 13, 2007 at 12:38:30AM +0200, Vortechz Anderson wrote: wine-0.9.37 compiles on OpenBSD 4.1, except the dnsapi Not for me. Do you have patches? The dnsapi error is this one, right? cc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_REENTRANT -fPIC -Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith -O2 -pipe -o name.o name.c In file included from name.c:46: dnsapi.h:143: error: syntax error before ns_msg dnsapi.h:144: error: syntax error before '*' token I could get this to compile by adding a LIB_DEPENDS on libbind and adding CFLAGS=-I/usr/local/include/bind But then I hit this error: cc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_REENTRANT -fPIC - Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith -I/usr/local/incl ude/bind -o ipstats.o ipstats.c ipstats.c: In function `getNumRoutes': ipstats.c:630: error: `RTF_MULTICAST' undeclared (first use in this function) ipstats.c:630: error: (Each undeclared identifier is reported only once ipstats.c:630: error: for each function it appears in.) ipstats.c: In function `getRouteTable': ipstats.c:704: error: `RTF_MULTICAST' undeclared (first use in this function) Two places in this file he's looking at the routing table using sysctl and wishes to leave out entries that are not gateways or that are multicast. RTF_MULTICAST is defined in route.h in freebsd and linux but not in OpenBSD. Looking at this paragraph from the netstat man page, I though perhaps this wouldn't be an issue for OpenBSD, maybe the multicast check could be left out, maybe that's a separate table, but I'm uncertain: -g Show information related to multicast (group address) routing. By default, show the IP multicast virtual-interface and routing tables. If the -s option is also present, show multicast routing statistics. I left it out for myself and then hit this error... cc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_NTSYSTEM_ -D_REENT RANT -fPIC -Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith -I/us r/local/include/bind -o cdrom.o cdrom.c cdrom.c: In function `DVD_ReadStructure': cdrom.c:1911: error: syntax error before s cdrom.c:1919: error: `s' undeclared (first use in this function) c This was a case where Linux, NetBSD, etc. have a typedef for the dvd_struct union in sys/cdio.h but OpenBSD has a regular union definition. Then I hit this... cc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_NTSYSTEM_ -D_REENT RANT -fPIC -Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith -I/us r/local/include/bind -o signal_i386.o signal_i386.c signal_i386.c:325: error: `T_MCHK' undeclared here (not in a function) signal_i386.c:325: error: enumerator value for `TRAP_x86_MCHK' not integer constan t signal_i386.c:349: error: `T_XMMFLT' undeclared here (not in a function) signal_i386.c:349: error: enumerator value for `TRAP_x86_CACHEFLT' not integer con stant signal_i386.c: In function `segv_handler': signal_i386.c:1186: error: duplicate case value signal_i386.c:1181: error: previously used here Here, although there's an if defined including __OpenBSD__ at the top of an enum, there are two definitions with slightly different names in OpenBSD trap.h than they try to use. I'm guessing there will be more stuff to come, but I was wondering if you already worked through all these yourself to get it to compile. I hope to get it to compile sometime this week, although I'm not sure if I'll be able to help with threading issues. The archives mention something about using vfork in a weird, linux specific way. Maybe that would be something to look at? If you look at .../libs/wine/mmap.c there are some troubling comments above and within try_mmap_fixed, as well as an #if defined(__NetBSD__). Execution leads to segfault. (Note: Generic kernel _has_ SYSV MSG/SHM/SEMand I have not forgot sysctl machdep.userldt=1) I know there are some issues about wine's use of kernel threads on OpenBSD. I am clueless about the true problem though. If possible, I would like some comments on the ktrace kdump. // V.A. 25268 ktrace RET ktrace 0 25268 ktrace CALL ... execve(0xcfbc4960,0xcfbc4ebc,0xcfbc4ec8) 25268 ktrace NAMI /usr/local/bin/wine 25268 wine NAMI /usr/libexec/ld.so 25268 wine EMUL native 25268 wine RET execve 0 25268 wine CALL issetugid() 25268 wine RET issetugid 0 25268
Re: wine-0.9.37 ktrace
On Sun, May 13, 2007 at 12:38:30AM +0200, Vortechz Anderson wrote: wine-0.9.37 compiles on OpenBSD 4.1, except the dnsapi Execution leads to segfault. (Note: Generic kernel _has_ SYSV MSG/SHM/SEMand I have not forgot sysctl machdep.userldt=1) I know there are some issues about wine's use of kernel threads on OpenBSD. I am clueless about the true problem though. If possible, I would like some comments on the ktrace kdump. looks familiar. before you even get to problems with threads, you have problems with wine wanting to control where things are located. I've got a port of 0.9.10 that gets a little farther than what you got here. loads libwine and libc, but cannot load ntdll.dll.so. try setting 'ac_cv_cflags__Wl___section_start__interp_0x7bf00400=no' in your environment before running configure and see if that gets you any farther. definitely not a trivial port. // V.A. 25268 ktrace RET ktrace 0 25268 ktrace CALL execve(0xcfbc4960,0xcfbc4ebc,0xcfbc4ec8) 25268 ktrace NAMI /bin/wine 25268 ktrace RET execve -1 errno 2 No such file or directory 25268 ktrace CALL execve(0xcfbc4960,0xcfbc4ebc,0xcfbc4ec8) 25268 ktrace NAMI /sbin/wine 25268 ktrace RET execve -1 errno 2 No such file or directory 25268 ktrace CALL execve(0xcfbc4960,0xcfbc4ebc,0xcfbc4ec8) 25268 ktrace NAMI /usr/bin/wine 25268 ktrace RET execve -1 errno 2 No such file or directory 25268 ktrace CALL execve(0xcfbc4960,0xcfbc4ebc,0xcfbc4ec8) 25268 ktrace NAMI /usr/sbin/wine 25268 ktrace RET execve -1 errno 2 No such file or directory 25268 ktrace CALL execve(0xcfbc4960,0xcfbc4ebc,0xcfbc4ec8) 25268 ktrace NAMI /usr/local/bin/wine 25268 wine NAMI /usr/libexec/ld.so 25268 wine EMUL native 25268 wine RET execve 0 25268 wine CALL issetugid() 25268 wine RET issetugid 0 25268 wine CALL mprotect(0x2a6d5000,0x1000,0x1) 25268 wine RET mprotect 0 25268 wine CALL mmap(0,0x1000,0x3,0x1002,0x,0,0,0) 25268 wine RET mmap -1 errno 12 Cannot allocate memory 25268 wine PSIG SIGSEGV SIG_DFL code 1 addr=0xa6d7bec trapno=1 25268 wine NAMI wine.core _ Flyger tiden iv?g? F?nga dagen med Yahoo! Mails inbyggda kalender. Dessutom 250 MB gratis, virusscanning och antispam. F? den p?: http://se.mail.yahoo.com -- [EMAIL PROTECTED] SDF Public Access UNIX System - http://sdf.lonestar.org
Re: wine-0.9.37 ktrace
Multireply, really... Mikolaj: RTHREADS in kernel makes no difference at this point and the problem seems to be within ld.so, see below... wine-0.9.37 compiles on OpenBSD 4.1, except the dnsapi Michael: I don't care whether there are dlls which does not compile until wine is running, but feel free to patch dnsapi if you really want to. Execution leads to segfault. (Note: Generic kernel _has_ SYSV MSG/SHM/SEMand I have not forgot sysctl machdep.userldt=1) I know there are some issues about wine's use of kernel threads on OpenBSD. I am clueless about the true problem though. If possible, I would like some comments on the ktrace kdump. looks familiar. before you even get to problems with threads, you have problems with wine wanting to control where things are located. I've got a port of 0.9.10 that gets a little farther than what you got here. loads libwine and libc, but cannot load ntdll.dll.so. try setting 'ac_cv_cflags__Wl___section_start__interp_0x7bf00400=no' in your environment before running configure and see if that gets you any farther. Uhmnot really definitely not a trivial port. I found this, so I think one has to figure out whether ld.so works as it should, before even thinking about a wine port. I have been really discouraged by now, so don't expect any fireworks. http://www.kerneltraffic.org/wine/wn20030829_185.html#3 [...cut...] -- View this message in context: http://www.nabble.com/wine-0.9.37-ktrace-tf3733527.html#a10629252 Sent from the openbsd user - ports mailing list archive at Nabble.com.
Re: wine-0.9.37 ktrace
Uh, well...some progress coming up...but I don't know whether I'm getting closer or just being lucky ;) (I'm sorry if this post is unclear and/or of very little use. I'm not used to twisting/tweaking/porting software.) I put: 'ac_cv_cflags__Wl___section_start__interp_0x7bf00400=no' exported that as well, and ran configure --disable-win16 I believe the win16 might be responsible for the DOS low mem mapping sc***ing things up...but I haven't checked that, just to confuse myself and everyone else... I commented out the dnsapi line in dlls/Makefile... next I compiled: make depend ; make and tried running: ktrace wine-pthreads /emul/w/windows/sol.exe (Note: I try that because sol.exe runs under the current wine-99 OpenBSD port.) ...which created a *huge* ktrace.out file output from program was only wine: failed to initialize: File not found Here are the last four rows of code in main.c, which is the only place I can find that error msg: wine_pthread_set_functions( pthread_functions, sizeof(pthread_functions) ); wine_init( argc, argv, error, sizeof(error) ); fprintf( stderr, wine: failed to initialize: %s\n, error ); exit(1); After kdumping into text file, I did some greps for relevant things: grep -n '.so' wtf 4: 20251 wine-pthread NAMI /usr/libexec/ld.so 18: 20251 wine-pthread NAMI /var/run/ld.so.hints 30: 20251 wine-pthread NAMI /usr/local/lib/libwine.so.1.0 226: 20251 wine-pthread NAMI /usr/lib/libpthread.so.7.0 392: 20251 wine-pthread NAMI /usr/lib/libc.so.40.3 571: 20251 wine-pthread NAMI /usr/lib/libossaudio.so.3.0 grep -n not found wtf 100890: wine: failed to initialize: File not found About 150 lines above, /usr/local/lib/wine is opened, so there might be something missing there...? The closest access I can see is this: (note: 20251 is pid) 20251 wine-pthread NAMI /usr/local/lib/wine 20251 wine-pthread RET open 5 20251 wine-pthread CALL fstat(0x5,0xcfbc8420) grep -n 'lib' wtf 4: 20251 wine-pthread NAMI /usr/libexec/ld.so 27: 20251 wine-pthread NAMI $ORIGIN/../lib 30: 20251 wine-pthread NAMI /usr/local/lib/libwine.so.1.0 223: 20251 wine-pthread NAMI $ORIGIN/../lib 226: 20251 wine-pthread NAMI /usr/lib/libpthread.so.7.0 389: 20251 wine-pthread NAMI $ORIGIN/../lib 392: 20251 wine-pthread NAMI /usr/lib/libc.so.40.3 568: 20251 wine-pthread NAMI $ORIGIN/../lib 571: 20251 wine-pthread NAMI /usr/lib/libossaudio.so.3.0 1624: 20251 wine-pthread NAMI /usr/local/lib/../bin/wineserver 100687: 20251 wine-pthread NAMI /usr/local/lib/wine 100702: 20251 wine-pthread NAMI /usr/local/lib/wine 100723: 20251 wine-pthread NAMI /usr/local/lib/wine 100736: 20251 wine-pthread NAMI /usr/local/lib/wine Uh..$ORIGIN?? Another grep, which I won't paste here, is allocate memory, which seems to occupy every other line from ~2400 to ~100600. Example: wine-pthread CALL mmap(0x9f23,0x13f,0,0x1442,0x,0,0,0) wine-pthread RET mmap -1 errno 12 Cannot allocate memory I don't understand why mmap is called with eight args, maybe it's trivial for those who know. // V.A. -- View this message in context: http://www.nabble.com/wine-0.9.37-ktrace-tf3733527.html#a10633012 Sent from the openbsd user - ports mailing list archive at Nabble.com.
Re: wine-0.9.37 ktrace
Vortechz wrote: [...cut...] grep -n not found wtf 100890: wine: failed to initialize: File not found About 150 lines above, /usr/local/lib/wine is opened, so there might be something missing there...? The closest access I can see is this: (note: 20251 is pid) 20251 wine-pthread NAMI /usr/local/lib/wine 20251 wine-pthread RET open 5 20251 wine-pthread CALL fstat(0x5,0xcfbc8420) NOTE: I did run make install as well, although I forgot to mention that. -- View this message in context: http://www.nabble.com/wine-0.9.37-ktrace-tf3733527.html#a10633282 Sent from the openbsd user - ports mailing list archive at Nabble.com.
Re: wine-0.9.37 ktrace
On Sun, May 13, 2007 at 12:38:30AM +0200, Vortechz Anderson wrote: wine-0.9.37 compiles on OpenBSD 4.1, except the dnsapi Not for me. Do you have patches? The dnsapi error is this one, right? cc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_REENTRANT -fPIC -Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith -O2 -pipe -o name.o name.c In file included from name.c:46: dnsapi.h:143: error: syntax error before ns_msg dnsapi.h:144: error: syntax error before '*' token I could get this to compile by adding a LIB_DEPENDS on libbind and adding CFLAGS=-I/usr/local/include/bind But then I hit this error: cc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_REENTRANT -fPIC - Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith -I/usr/local/incl ude/bind -o ipstats.o ipstats.c ipstats.c: In function `getNumRoutes': ipstats.c:630: error: `RTF_MULTICAST' undeclared (first use in this function) ipstats.c:630: error: (Each undeclared identifier is reported only once ipstats.c:630: error: for each function it appears in.) ipstats.c: In function `getRouteTable': ipstats.c:704: error: `RTF_MULTICAST' undeclared (first use in this function) Two places in this file he's looking at the routing table using sysctl and wishes to leave out entries that are not gateways or that are multicast. RTF_MULTICAST is defined in route.h in freebsd and linux but not in OpenBSD. Looking at this paragraph from the netstat man page, I though perhaps this wouldn't be an issue for OpenBSD, maybe the multicast check could be left out, maybe that's a separate table, but I'm uncertain: -g Show information related to multicast (group address) routing. By default, show the IP multicast virtual-interface and routing tables. If the -s option is also present, show multicast routing statistics. I left it out for myself and then hit this error... cc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_NTSYSTEM_ -D_REENT RANT -fPIC -Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith -I/us r/local/include/bind -o cdrom.o cdrom.c cdrom.c: In function `DVD_ReadStructure': cdrom.c:1911: error: syntax error before s cdrom.c:1919: error: `s' undeclared (first use in this function) c This was a case where Linux, NetBSD, etc. have a typedef for the dvd_struct union in sys/cdio.h but OpenBSD has a regular union definition. Then I hit this... cc -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_NTSYSTEM_ -D_REENT RANT -fPIC -Wall -pipe -fno-strict-aliasing -Wwrite-strings -Wpointer-arith -I/us r/local/include/bind -o signal_i386.o signal_i386.c signal_i386.c:325: error: `T_MCHK' undeclared here (not in a function) signal_i386.c:325: error: enumerator value for `TRAP_x86_MCHK' not integer constan t signal_i386.c:349: error: `T_XMMFLT' undeclared here (not in a function) signal_i386.c:349: error: enumerator value for `TRAP_x86_CACHEFLT' not integer con stant signal_i386.c: In function `segv_handler': signal_i386.c:1186: error: duplicate case value signal_i386.c:1181: error: previously used here Here, although there's an if defined including __OpenBSD__ at the top of an enum, there are two definitions with slightly different names in OpenBSD trap.h than they try to use. I'm guessing there will be more stuff to come, but I was wondering if you already worked through all these yourself to get it to compile. I hope to get it to compile sometime this week, although I'm not sure if I'll be able to help with threading issues. The archives mention something about using vfork in a weird, linux specific way. Maybe that would be something to look at? If you look at .../libs/wine/mmap.c there are some troubling comments above and within try_mmap_fixed, as well as an #if defined(__NetBSD__). Execution leads to segfault. (Note: Generic kernel _has_ SYSV MSG/SHM/SEMand I have not forgot sysctl machdep.userldt=1) I know there are some issues about wine's use of kernel threads on OpenBSD. I am clueless about the true problem though. If possible, I would like some comments on the ktrace kdump. // V.A. 25268 ktrace RET ktrace 0 25268 ktrace CALL ... execve(0xcfbc4960,0xcfbc4ebc,0xcfbc4ec8) 25268 ktrace NAMI /usr/local/bin/wine 25268 wine NAMI /usr/libexec/ld.so 25268 wine EMUL native 25268 wine RET execve 0 25268 wine CALL issetugid() 25268 wine RET issetugid 0 25268 wine CALL mprotect(0x2a6d5000,0x1000,0x1) 25268 wine RET mprotect 0 25268 wine CALL mmap(0,0x1000,0x3,0x1002,0x,0,0,0) 25268 wine RET mmap -1 errno 12 Cannot allocate memory 25268 wine PSIG SIGSEGV SIG_DFL code 1 addr=0xa6d7bec trapno=1 25268 wine NAMI wine.core -- Mike Small [EMAIL PROTECTED]
Re: wine-0.9.37 ktrace
Hi, On Sun, May 13, 2007 at 12:38:30AM +0200, Vortechz Anderson wrote: wine-0.9.37 compiles on OpenBSD 4.1, except the dnsapi Execution leads to segfault. (Note: Generic kernel _has_ SYSV MSG/SHM/SEMand I have not forgot sysctl machdep.userldt=1) I know there are some issues about wine's use of kernel threads on OpenBSD. I am clueless about the true problem though. If possible, I would like some comments on the ktrace kdump. Is it make sense to compile kernel with RTHREADS defined and try run wine? -- best regards q#
