[breakage] Set the correct permissions on the PE image sections.

[Uwe Bonnes]

Hallo,

the patch http://cvs.winehq.com/patch.py?id=1027987166850574216879071
"Set the correct permissions on the PE image sections"

breaks the propsheet handling on the xilinx webpack suite. When a propsheet
should be opened, appended backtrace  appears.

PROPSHEET_CreatePage writes to ppshpage->u.pResource when PSP_DLGINDIRECT is
set. MSDN also tells that applications will break if pResource is  not
writable. It seems that above patch maps the resource section write
protected.

Info segments doesn't tell about a range including 0x00467390, info maps
tells about VirtualQueryEx Unsupported .

Bye

[Alexandre Julliard]

Uwe Bonnes  writes:
> PROPSHEET_CreatePage writes to ppshpage->u.pResource when PSP_DLGINDIRECT is
> set. MSDN also tells that applications will break if pResource is  not
> writable. It seems that above patch maps the resource section write
> protected.

Actually it should be read-only, and the default exception handler is
supposed to unprotect it to hide bugs in broken apps. But writing to
resources is a bug in any case, and that propsheet code needs to be
fixed. In fact there's already a hack in there to avoid the same issue
with builtin resources; I guess it's really time to fix it properly.

[Uwe Bonnes]

> > > > > "Alexandre" == Alexandre Julliard  writes:

Alexandre> Uwe Bonnes  writes:
    >> PROPSHEET_CreatePage writes to ppshpage->u.pResource when
    >> PSP_DLGINDIRECT is set. MSDN also tells that applications will break
    >> if pResource is not writable. It seems that above patch maps the
    >> resource section write protected.

Alexandre> Actually it should be read-only, and the default exception
    Alexandre> handler is supposed to unprotect it to hide bugs in broken
    Alexandre> apps. But writing to resources is a bug in any case, and that
    Alexandre> propsheet code needs to be fixed. In fact there's already a
    Alexandre> hack in there to avoid the same issue with builtin resources;
    Alexandre> I guess it's really time to fix it properly.

Yes, I forget to note that it works with (NT) native commctrl and so the
protection seems right...

Bye

[Eric Pouech]

> Info segments doesn't tell about a range including 0x00467390, info maps
> tells about VirtualQueryEx Unsupported .

there's a hacky implemntation for VQEx floating around... 
if you need it grab it from:
http://www.winehq.com/hypermail/wine-devel/2002/07/0235.html

[Uwe Bonnes]

> > > > > "Eric" == Eric Pouech  writes:

>> Info segments doesn't tell about a range including 0x00467390, info
    >> maps tells about VirtualQueryEx Unsupported .
    Eric> there's a hacky implemntation for VQEx floating around...  if you
    Eric> need it grab it from:
    Eric> http://www.winehq.com/hypermail/wine-devel/2002/07/0235.html

What is so "hacky" about this patch?

[Eric Pouech]

Uwe Bonnes a écrit :
> > > > > > "Eric" == Eric Pouech  writes:
> >> Info segments doesn't tell about a range including 0x00467390, info
> >> maps tells about VirtualQueryEx Unsupported .
> Eric> there's a hacky implemntation for VQEx floating around...  if you
> Eric> need it grab it from:
> Eric> http://www.winehq.com/hypermail/wine-devel/2002/07/0235.html
>
> What is so "hacky" about this patch?

- linux only patch
- duplication of information between the information stored in process
address space vs server side (as well as returned values between VQ and
VQEx)
- ...
A+