January 2003 - Wine-Devel - list.winehq.org

Re: wine/programs/wineboot wineboot.c Makefile.in

by Dimitrie O. Paun

On January 11, 2003 05:26 pm, Shachar Shemesh wrote: > The problem is that buffer is defined as a > "const WCHAR *", while free expects "void *". Change the var to WCHAR *, why would you free a const string without a warning? -- Dimi.

Jan. 11, 2003

Re: [D3D8 - 1] resend of Jason Texture work + many fixes

by Raphaël Junqueira

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Saturday 11 January 2003 16:28, Sylvain Petreolle wrote: > --- Raphaël Junqueira <fenix(a)club-internet.fr> a écrit : > -----BEGIN > PGP SIGNED MESSAGE----- > > > Hash: SHA1 > > > > On Saturday 11 January 2003 01:27, Raphaël Junqueira wrote: > > > Hi again, > > > > I forgot to say with this patch now: > > > > - - i have run successfully the NVIDIA Effects Browser 3.0 > > (http://developer.nvidia.com/view.asp?IO=NVEffectsBrowser_3) > > with many effects as Aniso > > (http://developer.nvidia.com/view.asp?IO=Anisotropic_Lighting) or > > membrane_lighting > > (http://developer.nvidia.com/view.asp?IO=membrane_lighting) in pure > > vertex shader software emulation > > - - Warcraf3 Demo begin to run > > - - SplinterCell Demo now complain about stencil support > > > > anyone have other Dx8 games demos for testing (simple if possible) ? > > Re-Volt demo, wich made me dig into ddraw/dsurface code. <g> > http://www.re-volt.com/images/revoltdemo.zip hmm, it it use ddraw code, it use d3d version 1 to 7 and not 8 ;( so i can't help you with your surface problem Raphael > > ===== > Sylvain Petreolle > spetreolle(a)users.sourceforge.net > Fight against Spam ! http://www.euro.cauce.org/en/index.html > ICQ #170597259 > > "Don't think you are. Know you are." Morpheus, in "Matrix". > > ___________________________________________________________ > Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français ! > Yahoo! Mail : http://fr.mail.yahoo.com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (GNU/Linux) iD8DBQE+IGvCp7NA3AmQTU4RAqSgAJ42parsO5m4P1f/NGI23lWNkcmALwCghisS aLmCe6J2Bja6E6z1iNUmjCA= =v4z9 -----END PGP SIGNATURE-----

Jan. 11, 2003

C++ and winedbg: testers wanted

by Eric Pouech

as some folks have already noticed, loading ELF modules with C++ compilation units gave a zillion of traces while parsing debug info in winedbg this patch should silence this type of output it won't however add the C++ ABI to the debugger: heavy C++ debugging will require gdb (and also DwarfII debug info; stabs is so broken in this respect) on the C++ front, what should be possible is: - every class should be seen as a regular structures - non virtual methods will be available as functions (of course, this will be seen as a regular parameter, and functions will be seen as undecorated names... this may recall to some of us the old good days of cfront ;-) anyway, as gcc has some discrepencies it its stabs output in its version, I'd like to get some feedback of people loading C++ info into winedbg I already tested gcc 3.2 and 2.96, but results from 2.95, 3.0 and 3.3 would be appreciated TIA A+ -- Eric Pouech Name: wd_stabs ChangeLog: - added experimental parsing for C++ code (but winedbg doesn't support C++ ABI) - now loading symbols for included files License: X11 GenDate: 2003/01/11 17:14:34 UTC ModifiedFiles: programs/winedbg/stabs.c AddedFiles: =================================================================== RCS file: /home/cvs/cvsroot/wine/wine/programs/winedbg/stabs.c,v retrieving revision 1.3 diff -u -u -r1.3 stabs.c --- programs/winedbg/stabs.c 29 Sep 2002 18:35:31 -0000 1.3 +++ programs/winedbg/stabs.c 11 Jan 2003 17:14:30 -0000 @@ -31,6 +31,8 @@ * available (hopefully) from http:\\sources.redhat.com\gdb\onlinedocs */ +#define WITH_CPP_STABS_PARSING 1 + #include "config.h" #include <sys/types.h> @@ -269,13 +271,13 @@ if (**x=='(') { (*x)++; /* '(' */ - filenr=strtol(*x,x,10); /* <int> */ + filenr=strtol(*x,x,10); /* <int> */ (*x)++; /* ',' */ - subnr=strtol(*x,x,10); /* <int> */ + subnr=strtol(*x,x,10); /* <int> */ (*x)++; /* ')' */ } else { filenr = 0; - subnr = strtol(*x,x,10); /* <int> */ + subnr = strtol(*x,x,10); /* <int> */ } return DEBUG_FileSubNr2StabEnum(filenr,subnr); } @@ -343,14 +345,58 @@ if (*ptd->ptr++ != ';') return -1; /* ';' */ return 0; } +#ifdef WITH_CPP_STABS_PARSING +static inline int DEBUG_PTS_ReadMethodInfo(struct ParseTypedefData* ptd, + struct datatype* dt) +{ + char* tmp; + + if (*ptd->ptr++ == ':') + { + if (!(tmp = strchr(ptd->ptr, ';'))) return -1; + ptd->ptr = tmp + 1; + + } + if (!(*ptd->ptr >= '0' && *ptd->ptr <= '9')) return -1; + ptd->ptr++; + if (!(ptd->ptr[0] >= 'A' && *ptd->ptr <= 'D')) return -1; + ptd->ptr++; + if (*ptd->ptr++ != '.') return -1; + if (*ptd->ptr != ';') + { + while (*ptd->ptr) + { + /* some GCC versions (2.96 for example) really spit out wrongly defined + * stabs... in some cases, the methods are not correctly handled + * so just swallow the garbage until either .; (end of method) + * or ;; (end of struct) are found + */ + if (ptd->ptr[0] == '.' && ptd->ptr[1] == ';') + { + ptd->ptr++; + break; + } + if (ptd->ptr[0] == ';' && ptd->ptr[1] == ';') + break; + ptd->ptr++; + } + } + + return 0; +} +#endif -static inline int DEBUG_PTS_ReadAggregate(struct ParseTypedefData* ptd, struct datatype* sdt) +static inline int DEBUG_PTS_ReadAggregate(struct ParseTypedefData* ptd, + struct datatype* sdt) { int sz, ofs; char* last; struct datatype* adt; int idx; int doadd; +#ifdef WITH_CPP_STABS_PARSING + BOOL inMethod; +#endif sz = strtol(ptd->ptr, &last, 10); if (last == ptd->ptr) return -1; @@ -366,29 +412,67 @@ while (*ptd->ptr != ';') { /* agg_name : type ',' <int:offset> ',' <int:size> */ idx = ptd->idx; + if (DEBUG_PTS_ReadID(ptd) == -1) return -1; /* Ref. TSDF R2.130 Section 7.4. When the field name is a method name * it is followed by two colons rather than one. */ - if (*ptd->ptr == ':') ptd->ptr++; + if (*ptd->ptr == ':') + { + ptd->ptr++; +#ifdef WITH_CPP_STABS_PARSING + { + char* psemic = strchr(ptd->ptr, ';'); + char* psharp = strchr(ptd->ptr, '#'); + if (!psharp || psharp > psemic) + { + struct datatype* dt = NULL; + /* FIXME: hack... this shall be a ctor */ + if (DEBUG_PTS_ReadTypedef(ptd, NULL, &dt) == -1) return -1; + if (DEBUG_PTS_ReadMethodInfo(ptd, dt) == -1) return -1; + if (*ptd->ptr++ != ';') return -1; + ptd->idx = idx; + continue; + } + } + inMethod = TRUE; + } + else + { + inMethod = FALSE; +#endif + } + if (DEBUG_PTS_ReadTypedef(ptd, NULL, &adt) == -1) return -1; if (!adt) return -1; - if (*ptd->ptr++ != ',') return -1; - if (DEBUG_PTS_ReadNum(ptd, &ofs) == -1) return -1; - if (*ptd->ptr++ != ',') return -1; - if (DEBUG_PTS_ReadNum(ptd, &sz) == -1) return -1; - if (*ptd->ptr++ != ';') return -1; + if (*ptd->ptr != ',') + { +#ifdef WITH_CPP_STABS_PARSING + if (!inMethod) +#endif + return -1; + ptd->ptr++; + } + else + { + ptd->ptr++; + if (DEBUG_PTS_ReadNum(ptd, &ofs) == -1) return -1; + if (*ptd->ptr++ != ',') return -1; + if (DEBUG_PTS_ReadNum(ptd, &sz) == -1) return -1; + if (*ptd->ptr++ != ';') return -1; - if (doadd) DEBUG_AddStructElement(sdt, ptd->buf + idx, adt, ofs, sz); + if (doadd) DEBUG_AddStructElement(sdt, ptd->buf + idx, adt, ofs, sz); + } ptd->idx = idx; } ptd->ptr++; /* ';' */ return 0; } -static inline int DEBUG_PTS_ReadEnum(struct ParseTypedefData* ptd, struct datatype* edt) +static inline int DEBUG_PTS_ReadEnum(struct ParseTypedefData* ptd, + struct datatype* edt) { int ofs; int idx; @@ -405,7 +489,8 @@ return 0; } -static inline int DEBUG_PTS_ReadArray(struct ParseTypedefData* ptd, struct datatype* adt) +static inline int DEBUG_PTS_ReadArray(struct ParseTypedefData* ptd, + struct datatype* adt) { int lo, hi; struct datatype* rdt; @@ -463,14 +548,17 @@ /* then the real definitions */ switch (*ptd->ptr++) { case '*': +#ifdef WITH_CPP_STABS_PARSING + case '&': +#endif new_dt = DEBUG_NewDataType(DT_POINTER, NULL); if (DEBUG_PTS_ReadTypedef(ptd, NULL, &ref_dt) == -1) return -1; DEBUG_SetPointerType(new_dt, ref_dt); break; - case 'k': /* 'const' modifier */ - case 'B': /* 'volatile' modifier */ - /* just kinda ignore the modifier, I guess -gmt */ - if (DEBUG_PTS_ReadTypedef(ptd, NULL, &new_dt) == -1) return -1; + case 'k': /* 'const' modifier */ + case 'B': /* 'volatile' modifier */ + /* just kinda ignore the modifier, I guess -gmt */ + if (DEBUG_PTS_ReadTypedef(ptd, NULL, &new_dt) == -1) return -1; break; case '(': ptd->ptr--; @@ -588,7 +676,8 @@ case 31: basic = DT_BASIC_LONGLONGINT; break; case 32: basic = DT_BASIC_ULONGLONGINT; break; default: - DEBUG_Printf(DBG_CHN_MESG, "Unsupported integral type (%d/%d)\n", lo, sz); + DEBUG_Printf(DBG_CHN_MESG, + "Unsupported integral type (%d/%d)\n", lo, sz); return -1; } if (!(new_dt = DEBUG_GetBasicType(basic))) { @@ -598,6 +687,35 @@ if (*ptd->ptr++ != ';') return -1; } break; +#ifdef WITH_CPP_STABS_PARSING + case '#': + new_dt = DEBUG_NewDataType(DT_FUNC, NULL); + if (*ptd->ptr == '#') + { + ptd->ptr++; + if (DEBUG_PTS_ReadTypedef(ptd, NULL, &ref_dt) == -1) return -1; + DEBUG_SetPointerType(new_dt, ref_dt); + if (*ptd->ptr++ != ';') return -1; + + } + else + { + struct datatype* cls_dt; + struct datatype* pmt_dt; + + if (DEBUG_PTS_ReadTypedef(ptd, NULL, &cls_dt) == -1) return -1; + if (*ptd->ptr++ != ',') return -1; + if (DEBUG_PTS_ReadTypedef(ptd, NULL, &ref_dt) == -1) return -1; + DEBUG_SetPointerType(new_dt, ref_dt); + while (*ptd->ptr++ == ',') + { + if (DEBUG_PTS_ReadTypedef(ptd, NULL, &pmt_dt) == -1) return -1; + } + } + if (DEBUG_PTS_ReadMethodInfo(ptd, new_dt) == -1) return -1; + + break; +#endif default: DEBUG_Printf(DBG_CHN_MESG, "Unknown type '%c'\n", ptd->ptr[-1]); return -1; @@ -647,11 +765,15 @@ if ((ptd.ptr = strchr(ptr, ':'))) { ptd.ptr++; if (*ptd.ptr != '(') ptd.ptr++; +#ifdef WITH_CPP_STABS_PARSING + /* most of type definitions take one char, except Tt */ + if (*ptd.ptr != '(') ptd.ptr++; +#endif ret = DEBUG_PTS_ReadTypedef(&ptd, typename, &dt); } if (ret == -1 || *ptd.ptr) { - DEBUG_Printf(DBG_CHN_MESG, "failure on %s at %s\n", ptr, ptd.ptr); + DEBUG_Printf(DBG_CHN_MESG, "Failure on %s at %s\n", ptr, ptd.ptr); return FALSE; } @@ -729,7 +851,7 @@ strtabinc = 0; stabbuff[0] = '\0'; - for(i=0; i < nstab; i++, stab_ptr++ ) + for (i = 0; i < nstab; i++, stab_ptr++) { ptr = strs + (unsigned int) stab_ptr->n_un.n_name; if( ptr[strlen(ptr) - 1] == '\\' ) @@ -944,7 +1066,7 @@ } #endif - if( *ptr == '\0' ) + if( *ptr == '\0' ) /* end of N_SO file */ { /* * Nuke old path. @@ -1004,7 +1127,7 @@ stabbuff[0] = '\0'; #if 0 - DEBUG_Printf(DBG_CHN_MESG, "%d %x %s\n", stab_ptr->n_type, + DEBUG_Printf(DBG_CHN_MESG, "0x%02x %x %s\n", stab_ptr->n_type, (unsigned int) stab_ptr->n_value, strs + (unsigned int) stab_ptr->n_un.n_name); #endif @@ -1044,7 +1167,7 @@ nsym = symtab->sh_size / sizeof(*symp); strp = (char *) (addr + strtab->sh_offset); - for(i=0; i < nsym; i++, symp++) + for (i = 0; i < nsym; i++, symp++) { /* * Ignore certain types of entries which really aren't of that much @@ -1165,7 +1288,7 @@ } if (stabsect == -1 || stabstrsect == -1) { - DEBUG_Printf(DBG_CHN_WARN, "no .stab section\n"); + DEBUG_Printf(DBG_CHN_WARN, "No .stab section\n"); goto leave; } @@ -1181,7 +1304,7 @@ dil = DIL_LOADED; } else { dil = DIL_ERROR; - DEBUG_Printf(DBG_CHN_WARN, "bad stabs\n"); + DEBUG_Printf(DBG_CHN_WARN, "Couldn't read correctly read stabs\n"); goto leave; } @@ -1350,7 +1473,9 @@ if (filename == NULL) return DIL_ERROR; if (DEBUG_FindModuleByName(filename, DMT_ELF)) return DIL_LOADED; +#ifndef WITH_CPP_STABS_PARSING if (strstr (filename, "libstdc++")) return DIL_ERROR; /* We know we can't do it */ +#endif dil = DEBUG_ProcessElfFile(filename, load_offset, dyn_addr); /* if relative pathname, try some absolute base dirs */ @@ -1383,6 +1508,7 @@ for (lm_addr = (u_long)dbg_hdr->r_map; lm_addr; lm_addr = (u_long)lm.l_next) { if (!DEBUG_READ_MEM_VERBOSE((void*)lm_addr, &lm, sizeof(lm))) return FALSE; + if (lm.l_addr != 0 && DEBUG_READ_MEM_VERBOSE((void*)lm.l_addr, &ehdr, sizeof(ehdr)) && ehdr.e_type == ET_DYN && /* only look at dynamic modules */

Jan. 11, 2003

Re: patch for "GetFileTitleW: illegal call to GetFileTitleA"

by liu spider

--- Dmitry Timoshkov <dmitry(a)baikal.ru> wrote: > It would be a good idea to translate the result from > unicode to ANSI. > One more suggestion: > RtlCreateUnicodeStringFromAsciiz is supposed to > reduce trouble of conversion ASCIIZ strings to > unicode. Thanks for your clues, I will make it. liuspider __________________________________________________ Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now. http://mailplus.yahoo.com

Jan. 11, 2003

Redraw problems with Installshield 6

by Uwe Bonnes

Hallo, on and off there are reports of redraw problems. One class of applications that have redraw problems for me are Installshield 6 applications. When in a setup finally the installation begins, only the windows displaying the file(s) installed is updated, the rest of the IKERNEL.exe window isn't updated and so displays a lot of visual artefacts. Bye -- Uwe Bonnes bon(a)elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------

Jan. 11, 2003

Re: [D3D8 - 1] resend of Jason Texture work + many fixes

by Sylvain Petreolle

--- Raphaël Junqueira <fenix(a)club-internet.fr> a écrit : > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On Saturday 11 January 2003 01:27, Raphaël Junqueira wrote: > > Hi again, > > I forgot to say with this patch now: > > - - i have run successfully the NVIDIA Effects Browser 3.0 > (http://developer.nvidia.com/view.asp?IO=NVEffectsBrowser_3) > with many effects as Aniso > (http://developer.nvidia.com/view.asp?IO=Anisotropic_Lighting) or > membrane_lighting > (http://developer.nvidia.com/view.asp?IO=membrane_lighting) in pure > vertex shader software emulation > - - Warcraf3 Demo begin to run > - - SplinterCell Demo now complain about stencil support > > anyone have other Dx8 games demos for testing (simple if possible) ? Re-Volt demo, wich made me dig into ddraw/dsurface code. <g> http://www.re-volt.com/images/revoltdemo.zip ===== Sylvain Petreolle spetreolle(a)users.sourceforge.net Fight against Spam ! http://www.euro.cauce.org/en/index.html ICQ #170597259 "Don't think you are. Know you are." Morpheus, in "Matrix". ___________________________________________________________ Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français ! Yahoo! Mail : http://fr.mail.yahoo.com

Jan. 11, 2003

Re: patch for "GetFileTitleW: illegal call to GetFileTitleA"

by Dmitry Timoshkov

"liu spider" <liuspider(a)yahoo.com> wrote: > this is a patch to get rid of W->A calls: > GetFileTitleW: illegal call to GetFileTitleA > + ret = GetFileTitleW(lpwstrFile, lpwstrTitle, cbBuf); > + HeapFree(GetProcessHeap(), 0, lpwstrTitle); > + HeapFree(GetProcessHeap(), 0, lpwstrFile); > + return ret; > +} It would be a good idea to translate the result from unicode to ANSI. One more suggestion: RtlCreateUnicodeStringFromAsciiz is supposed to reduce trouble of conversion ASCIIZ strings to unicode. -- Dmitry.

Jan. 11, 2003

Re: Some ddraw / GDI palette problem

by Christian Costa

Lionel Ulmer wrote: >Hi all, > >In 'old' D3D games, a lot of textures are created paletted... And each >of these texture has its own DDraw palette object. And thus each has >some GDI memory allocated. > >This means that after some time, we run out of GDI Heap space due to >all these palettes created there : > >trace:ddraw:Main_DirectDraw_CreatePalette (0x4038ff48)->(00000044,0x406b1340,0x406b1978,(nil)) >trace:palette:CreatePalette entries=256 >err:local:LOCAL_GetBlock not enough space in GDI heap 01b7 for 1052 bytes > >Now that problem is fixed in one game (it was due to some bad reference >counting is some D3D code fixed in patch 79), but if ever it happens in yet >another game, is it possible to increase this space ? > > Lionel > I have a similar problem in SS2 and the patch 79 does not help. :-( According to the code, the GDI heap is 64K at most. So it cannot be expanded. :-( 64K sound like 16-bit stuff, I wonder if 32-bit systems have this limitation. Experts in this area surely have more details about this.... :-) Christian

Jan. 11, 2003

Re: Wine, Mozilla & IE ?

by Dan Kegel

Shachar Shemesh wrote: > THE thing that makes me miss C++ the most is exception handling. C > offers you, basically, three options: > > 1. Having to perform an "if" every other line and perform a *shudder* > goto in case of error. > 2. Have nested "if"s seven levels deep. > 3. return from the middle of the function. > 4. Worst yet, write insecure code. > > 1. produces the longest code. It is also the ugliest one, structured > programming wise. Actually, the Linux kernel uses this technique, and it doesn't always lead to longer code than, say, nested if's. It can actually be quite clean, and given that the use of 'goto' is restricted to a very narrow idiom, it's not all that hard to read. > Exceptions, on the other hand, give a clean solution to the entire > problem. ... Or at least seems to. It's *very* difficult to write C++ code that is threadsafe and exception safe at the same time. > If the discussion goes forward, I can give an example of rewriting my > latest patch to wineboot (currently using approach 1) with the other > four approaches (2, 3, 4 and exceptions). I've enjoyed the bit of Wine work I've done, but I'm not sure I'd continue enjoying it if C++ began to be used inside the core of Wine. C++ is for applications, not operating systems. - Dan -- Dan Kegel Linux User #78045 http://www.kegel.com

Jan. 11, 2003

Re: Help with /?? stuff

by Dan Kegel

Shachar Shemesh wrote: > [string processing stuff deleted] > I was seriously considering doing it in C++ (which SHOULD be ok, as this > is not a DLL, right?). That would have allowed much more felxibility. Oh > well, maybe next time. You're joking, right? - Dan -- Dan Kegel Linux User #78045 http://www.kegel.com

Jan. 11, 2003