Re: Windows 9X crash (was: [PATCH] Override Windows default Win-* key combinations when using Emacs)

[Fabrice Popineau <fabrice.popineau@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1727 bytes --]

Hi,

Strange that HeapAlloc() returns 4-byte aligned blocks and that is not
documented.
I used to have MSDN cdroms from that time
I would write a small program doing random allocation
and checking the alignment of the blocks returned.
Just to make sure that the problem does not lie elsewhere (?)

Fabrice

2016-01-15 9:08 GMT+01:00 Eli Zaretskii <eliz@gnu.org>:

> (Adding Fabrice, who wrote the new allocator code, to the discussion.)
>
> > Cc: emacs-devel@gnu.org
> > From: Jussi Lahdenniemi <jussi@aprikoodi.fi>
> > Date: Fri, 15 Jan 2016 09:52:38 +0200
> >
> > On 15.1.2016 8.56, Jussi Lahdenniemi wrote:
> > >> The first thing I'd like to know is what buffer is
> > >> that (I'm guessing *scratch* or *Messages*), and what is the value of
> > >> 'a' in this call frame:
> > >
> > > the value of the name variable in the Fget_buffer_create function is "
> > > *load*", and a in frame #4 is 0x02a22101.
> >
> > Ah, of course; when allocating 'b' in Fget_buffer_create, Emacs got the
> > pointer 0x...fc, which is NOT aligned at an 8-byte boundary as required
> > by USE_LSB_TAG.  This messes up the tags and causes the crash.
> >
> > So, apparently, on Windows 98 HeapAlloc does not guarantee 8-byte
> > alignment of memory.  Not that this would be documented anywhere...
>
> Right, this can explain everything.  It probably means that a build
> configured --with-wide-int should also be tested there, as it has
> different alignment needs.
>
> > How should we fix this?  I can write and test the fix, but I'd like to
> > hear your opinion on the preferred mechanism.
>
> What fix did you have in mind?  Over-allocating and recording the
> offset in the initial part of the block that we don't pass to the
> application?
>

[-- Attachment #2: Type: text/html, Size: 2418 bytes --]