Re: FW: Re: [emacs_user@hotmail.com: ***MEMORY-ERROR***: emacs[5172]: GSlice: failed

[Jan DjÃrv <jan dot h dot d at swipnet dot se>]

Christopher Faylor wrote:

> On Mon, Feb 19, 2007 at 08:26:36AM +0100, Jan Dj??rv wrote:
> > Christopher Faylor wrote:
> > > If someone is positing that one of several functions possibly isn't
> > > working in cygwin why not report exactly which function that would be?
> > > I.e., a little more work than reposting speculation would be
> > > appreciated.
> >
> > I did that comment, it is not speculation.  I currently have no W32
> > machine, hence no cygwin at all.
>
> So if you have no way to verify anything then it sounds like it would be
> speculation by definition.  Speculation isn't a bad word but speculating
> doesn't actually do any good until someone can confirm the speculation.
> Since I doubt that anyone here is going to download emacs to satisfy a
> speculation that cygwin's API is broken, we obviously need more details
> from people who care about this.

The code in question in glib looks like this:

static gpointer
allocator_memalign (gsize alignment,
                    gsize memsize)
{
  gpointer aligned_memory = NULL;
  gint err = ENOMEM;
#if     HAVE_COMPLIANT_POSIX_MEMALIGN
  err = posix_memalign (&aligned_memory, alignment, memsize);
#elif   HAVE_MEMALIGN
  errno = 0;
  aligned_memory = memalign (alignment, memsize);
  err = errno;
#elif   HAVE_VALLOC
  errno = 0;
  aligned_memory = valloc (memsize);
  err = errno;
#else
  /* simplistic non-freeing page allocator */
  ...
#endif


The #else part only calls malloc, and I assumed that it works on cygwin.  It 
would be a small thing to figure out which part is used on cygwin if the 
config.h was available.

> i.e.:
>
> > I don't know how the original poster configured Gtk+, I don't know
> > which version of cygwin he/she has, I don't know which version of Gtk+
> > he/she has.  I only tried to find out if Emacs could fix this somehow,
> > which it can't.  If the original poster (emacs_user@hotmail.com) can
> > send in his/hers config.h from the Gtk+ configuration, we can figure
> > out what function we are talking about.
>
> So, basically you're saying "a little more work" is required.

Yes, checking config.h for glib.


	Jan D.


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/