June 2003 - Wine-devel - list.winehq.org

Memory Protection Wierdness
by Robert Lunnon 30 Jun '03

30 Jun '03

I am playing with Worms II and the installer that comes with the Demo and uncovered a signal handling fault Under Solaris and some wierdness in the installer. After decompressing the application the installer executes the setup program which displays a splash screen the application then generates a Page Fault Exception. I can't give a stack trace because here is where the wierdness is. The Setup Program seems to not inherit the -debugmsg status and doesn't end up under control of the debugger (By this time there are 5 wine processes running) Anyway to cut a long story short the page fault DOES occur in a mapped region (Its in VIRTUAL_DumpViews() listing) but doesn't have a handler associated with it. This makes VirtualHandleFault fall through to the following code BYTE vprot = view->prot[((char *)addr - (char *)view->base) >> page_shift]; void *page = (void *)((UINT_PTR)addr & ~page_mask); char *stack = (char *)NtCurrentTeb()->stack_base + SIGNAL_STACK_SIZE + page_mask + 1; if (vprot & VPROT_GUARD) { VIRTUAL_SetProt( view, page, page_mask + 1, vprot & ~VPROT_GUARD ); ret = STATUS_GUARD_PAGE_VIOLATION; } /* is it inside the stack guard pages? */ if (((char *)addr >= stack) && ((char *)addr < stack + 2*(page_mask+1))) ret = STATUS_STACK_OVERFLOW; But of course this code doesn't reset the page protections to allow the access. I Doctored this code so it would always allow the access and the installer then completed OK, But worms itself will not run yet Q. Should all views have handlers ? Under What Conditions would they not ???

2 1

Re: RESUBMIT: --disable-debug -> no -gstabs+
by Alexandre Julliard 30 Jun '03

30 Jun '03

Marcus Meissner <marcus(a)jet.franken.de> writes: > Changelog: > Except when we explicitly --disable-debug, add -gstabs+ > if possible. I don't see much point in doing that, --disable-debug is broken anyway (and should probably be removed). You need to tweak the CFLAGS to make it work right, and then you can just as well get rid of -gstabs+ in the CFLAGS too. -- Alexandre Julliard julliard(a)winehq.com

1 0

Official MS symbol sets
by Mike Hearn 30 Jun '03

30 Jun '03

Hi, http://www.microsoft.com/whdc/ddk/debugging/symbolpkg.mspx I assume people have looked at this before - how easy are they to match up with Wine, and do people use them for debugging faults in native code? thanks -mike -- Mike Hearn <m.hearn(a)signal.qinetiq.com> QinetiQ - Malvern Technology Center

2 1

Package setups
by Mike Hearn 30 Jun '03

30 Jun '03

hi, I'm just looking at the HOWTO and want to clarify that all the Wine RPMs/packages attempt to set up a fake windows drive and configuration for the user. I'm under the impression that binary packages should work out of the box, so the user has no need to create their own config or fake windows drive. Is this right? If so, I'll amend the HOWTO to suit.

1 0

difference between delivery delays introduced by list-server and web-archive
by Saulius Krasuckas 30 Jun '03

30 Jun '03

hello, newbie is testing delays. thank you for not paying your attention. good luck

1 0

regression of patch 8366
by Oleg Prokhorov 30 Jun '03

30 Jun '03

Hello , This patch http://cvs.winehq.com/patch.py?root=/home/winehq/opt/cvs-commit&id=8366 Caused regression with chess.net application. (http://www.chess.net/cnfw/download/cnfw254.exe) If you click on a help button it will drop a messagebox which tells that 'No browser found', it is normal. The anomality is that it is not possible to close messagebox by clicking "ok" button with a mouse. But both pressing "enter" and clicking close window icon work. The line which produce the problem is (dlls/ole2/compobj.c) apt->win = CreateWindowA(aptWinClass, NULL, 0, 0, 0, 0, 0, 0, 0, OLE32_hInstance, NULL); -- Best regards, Oleg mailto:xolegpro(a)rbcmail.ru

1 0

Re: Known Wine ToDo's
by Shachar Shemesh 30 Jun '03

30 Jun '03

Tom wrote: > Hello Shachar, > > I am going to update the ToDo page : > http://www.winehq.com/?page=status_todo > > And im wondering if you can send me a list of ToDo's that your > aware of ? > > Thanks > > Tom > The tasks I'm talking about here are long term. Not all of them have been discussed on wine-devel yet. I'm CCing wine-devel in case anyone wants to object. National Language Support - Remove "Hebrew and Arabic" - it's just part of BiDi if technology is what wer'e talking about. If interface localization is what wer'e talking about, then just add "add localization to more languages". As a subitem of keyboard support, add "allow using a keyboard with keys outside the current locale codepage". Tools - Add "perform Windows' reboot operations automatically when required". Under "Low priority" add "Windows ACL support". -- Shachar Shemesh Open Source integration consultant Home page & resume - http://www.shemesh.biz/

1 0

Re: [MSRLE32] Some bugs and improvements.
by Tom 29 Jun '03

29 Jun '03

Michael Günnewig wrote: >Changelog: > - Fixed interframe compression - now works for RLE4 and RLE8. > - Fixed alignment of RLE4 in absolute mode. > - Improved compression (smaller output). > > Michael Günnewig > > > >------------------------------------------------------------------------ > >Index: dlls/msvideo/msrle32/msrle32.c >=================================================================== >RCS file: /home/wine/wine/dlls/msvideo/msrle32/msrle32.c,v >retrieving revision 1.4 >diff -d -u -r1.4 msrle32.c >--- dlls/msvideo/msrle32/msrle32.c 13 Jun 2003 23:17:01 -0000 1.4 >+++ dlls/msvideo/msrle32/msrle32.c 29 Jun 2003 17:32:06 -0000 >@@ -1,5 +1,5 @@ > /* >- * Copyright 2002 Michael Günnewig >+ * Copyright 2002-2003 Michael Günnewig > * > * This library is free software; you can redistribute it and/or > * modify it under the terms of the GNU Lesser General Public >@@ -17,8 +17,7 @@ > */ > > /* TODO: >- * - compression of RLE4 is buggy -- see FIXME's >- * - many improvements possible >+ * - some improvements possible > * - implement DecompressSetPalette? -- does we need it for anything? > */ > >@@ -297,10 +296,13 @@ > if (ColorCmp(clr1, clr3) <= lDist && > ColorCmp(clr2, clr4) <= lDist) { > /* diff at end? -- look-ahead for atleast ?? more encodable pixel */ >- >- /* FIXME */ >- >- return count; >+ if (pos + 2 < width && ColorCmp(clr1,lpB[pos+1]) <= lDist && >+ ColorCmp(clr2,lpB[pos+2]) <= lDist) { >+ if (pos + 4 < width && ColorCmp(lpB[pos+1],lpB[pos+3]) <= lDist && >+ ColorCmp(lpB[pos+2],lpB[pos+4]) <= lDist) >+ return count - 3; /* followed by atleast 4 encodable pixels */ >+ return count - 2; >+ } > } else if (lpP != NULL && ColorCmp(lpP[pos], lpB[pos]) <= lDist) { > /* 'compare' with previous frame for end of diff */ > INT count2 = 0; >@@ -331,9 +333,11 @@ > > for (count = 0; pos < width; pos++, count++) { > if (ColorCmp(lpA[pos], lpB[pos]) <= lDist) { >- /* diff at end? -- look-ahead for atleast one more encodable pixel */ >- if (pos + 1 < width && ColorCmp(lpA[pos], lpB[pos+1]) <= lDist) >- return count; >+ /* diff at end? -- look-ahead for some more encodable pixel */ >+ if (pos + 1 < width && ColorCmp(lpB[pos], lpB[pos+1]) <= lDist) >+ return count - 1; >+ if (pos + 2 < width && ColorCmp(lpB[pos+1], lpB[pos+2]) <= lDist) >+ return count - 1; > } else if (lpP != NULL && ColorCmp(lpP[pos], lpB[pos]) <= lDist) { > /* 'compare' with previous frame for end of diff */ > INT count2 = 0; >@@ -382,6 +386,8 @@ > /* add some pixel for absoluting if possible */ > count += countDiffRLE4(lpP, lpC - 1, lpC, pos-1, lDist, lpbi->biWidth); > >+ assert(count > 0); >+ > /* check for near end of line */ > if (x + count > lpbi->biWidth) > count = lpbi->biWidth - x; >@@ -390,9 +396,11 @@ > while (count > 2) { > INT i; > INT size = min(count, 254); >- BOOL extra_byte = (size / 2) % 2 || (size % 2); >+ int bytes = ((size + 1) & (~1)) / 2; >+ BOOL extra_byte = bytes & 0x01; > >- *lpSizeImage += 2 + size/2 + (size % 2) + extra_byte; >+ *lpSizeImage += 2 + bytes + extra_byte; >+ assert(((*lpSizeImage) % 2) == 0); > count -= size; > *lpOut++ = 0; > *lpOut++ = size; >@@ -468,6 +476,8 @@ > /* add some more pixels for absoluting if possible */ > count += countDiffRLE8(lpP, lpC - 1, lpC, pos-1, lDist, lpbi->biWidth); > >+ assert(count > 0); >+ > /* check for over end of line */ > if (x + count > lpbi->biWidth) > count = lpbi->biWidth - x; >@@ -663,6 +673,10 @@ > assert(lpOut == lpOutStart + lpbiOut->biSizeImage); > } > } >+ >+ /* add EOL -- will be changed to EOI */ >+ lpbiOut->biSizeImage += 2; >+ *((LPWORD)lpOut)++ = 0; > } > > /* change EOL to EOI -- end of image */ >@@ -789,11 +803,14 @@ > if (jumpy == 0) { > /* add EOL -- end of line */ > lpbiOut->biSizeImage += 2; >- *lpOut++ = 0; >- *lpOut++ = 0; >+ *((LPDWORD)lpOut)++ = 0; > assert(lpOut == (lpOutStart + lpbiOut->biSizeImage)); > } > } >+ >+ /* add EOL -- will be changed to EOI */ >+ lpbiOut->biSizeImage += 2; >+ *((LPDWORD)lpOut)++ = 0; > } > > /* change EOL to EOI -- end of image */ >@@ -847,7 +864,7 @@ > } > break; > default: /* absolute mode */ >- extra_byte = (code1 / 2) & 0x01 || (code1 % 2); >+ extra_byte = (((code1 + 1) & (~1)) / 2) & 0x01; > > if (pixel_ptr/bytes_per_pixel + code1 > lpbi->biWidth) > return ICERR_ERROR; >@@ -1468,9 +1485,7 @@ > > if (lpic->lpbiOutput->biBitCount == 4) > MSRLE32_CompressRLE4(pi, lpic->lpbiInput, (LPBYTE)lpic->lpInput, >- lpic->lpbiOutput, (LPBYTE)lpic->lpOutput, TRUE); >- /*MSRLE32_CompressRLE4(pi, lpic->lpbiInput, (LPBYTE)lpic->lpInput, >- lpic->lpbiOutput, (LPBYTE)lpic->lpOutput, (lpic->dwFlags & ICCOMPRESS_KEYFRAME) != 0);*/ >+ lpic->lpbiOutput, (LPBYTE)lpic->lpOutput, (lpic->dwFlags & ICCOMPRESS_KEYFRAME) != 0); > else > MSRLE32_CompressRLE8(pi, lpic->lpbiInput, (LPBYTE)lpic->lpInput, > lpic->lpbiOutput, (LPBYTE)lpic->lpOutput, (lpic->dwFlags & ICCOMPRESS_KEYFRAME) != 0); > >

1 0

Re: Uninitialised variable in TASK_Create
by Saulius Krasuckas 29 Jun '03

29 Jun '03

> Hi, > > There is an unintialised string (the name variable) in TASK_Create in > task.c (visible if you run almost any program with --debugmsg +task). > I would try and fix the problem but I have no idea what the correct > fix is. Is this just a case of adding *name = 0? > > Rob hello, (heh, i've lost the race.. :P) cents: i have found patch, which made my valgrind to scream bout the same issue: http://cvs.winehq.com/patch.py?root=/home/winehq/opt/cvs-commit&id=8245 Index: wine/loader/task.c diff -u wine/loader/task.c:1.132 wine/loader/task.c:1.133 --- wine/loader/task.c:1.132 Sun Jun 29 20:22:43 2003 +++ wine/loader/task.c Sun Jun 29 20:22:43 2003 <skipped> @@ -299,13 +300,16 @@ /* Copy the module name */ - GetModuleName16( pModule->self, name, sizeof(name) ); - strncpy( pTask->module_name, name, sizeof(pTask->module_name) ); + if (hModule) + { + GetModuleName16( hModule, name, sizeof(name) ); + strncpy( pTask->module_name, name, sizeof(pTask->module_name) ); + } in addition to null-initializing the "name", i'd add this at several page-downs: - pTask->compat_flags = GetProfileIntA( "Compatibility", name, 0 ); + if(name) + { + pTask->compat_flags = GetProfileIntA( "Compatibility", name, 0 ); + } IMHO we'd "save" two api-calls (GetProfileIntA() ant GetPrivateProfileIntA()) using such check.

1 0

Uninitialised variable in TASK_Create
by Robert Shearman 29 Jun '03

29 Jun '03

Hi, There is an unintialised string (the name variable) in TASK_Create in task.c (visible if you run almost any program with --debugmsg +task). I would try and fix the problem but I have no idea what the correct fix is. Is this just a case of adding *name = 0? Rob

1 0