Re: Resubmitting patch for memory-mapping problem -- See VirtualQuery post in wine-devel

[Tony Lambregts]

Massimo wrote:
> Was this patch lost in space or does it have some problem ?
>
> Took off in executable mapping VPROT_WRITE and VPROT_WRITECOPY flags 
> and
> added VPROT_EXEC; that looks much more close to win2k mapping.
> The only difference now is in Type field, wine maps as MEM_PRIVATE and win2k
> as MEM_IMAGE (I guess that's the right one, but no time now to check.
>
> The patch solves (partially) bug 890, and allows AutoCAD 2000 to start up (only in 
> win95 mode, because of another problem on fibers).
>
> Here the attached patch.... any comment appreciated :-)
>
> Regards
>
> Max
>
> ------------------------------------------------------------------------
>
> Index: virtual.c
> ===================================================================
> RCS file: /home/wine/wine/memory/virtual.c,v
> retrieving revision 1.80
> diff -u -r1.80 virtual.c
> --- virtual.c	3 Jul 2002 21:10:44 -0000	1.80
> +++ virtual.c	19 Jul 2002 19:33:10 -0000
> @@ -94,7 +94,6 @@
>     PAGE_EXECUTE_WRITECOPY      /* READ | WRITE | EXEC | WRITECOPY */
> };
>
> -
> static FILE_VIEW *VIRTUAL_FirstView;
> static CRITICAL_SECTION csVirtual = CRITICAL_SECTION_INIT("csVirtual");
>
> @@ -345,6 +344,7 @@
> ) {
>     if (protect) {
>     	*protect = VIRTUAL_Win32Flags[vprot & 0x0f];
> +
> /*    	if (vprot & VPROT_GUARD) *protect |= PAGE_GUARD;*/
>     	if (vprot & VPROT_NOCACHE) *protect |= PAGE_NOCACHE;
>
> @@ -650,7 +650,8 @@
>
> if (removable) hmapping = 0;  /* don't keep handle open on removable media */
>     if (!(view = VIRTUAL_CreateView( ptr, total_size, 0,
> -                                     VPROT_COMMITTED|VPROT_READ|VPROT_WRITE|VPROT_WRITECOPY,

this next line is redundant and should not be in the patch
> +// PATCH                                     VPROT_COMMITTED|VPROT_READ|VPROT_WRITE|VPROT_WRITECOPY,
> +/*PATCH*/                                     VPROT_COMMITTED|VPROT_READ|VPROT_EXEC,
>                                      hmapping )))

the  /*patch */ should not be in the patch file.
> {
>         SetLastError( ERROR_OUTOFMEMORY );

Looks like Alexandre was at least thinking of this (a little) when he 
wrote this patch

http://cvs.winehq.com/patch.py?id=1027987166850574216879071

You should probably try this out and see if it fixes bug 890.  <g>


Tony Lambregts