Paul Eggert <eggert@cs.ucla.edu> writes:

> Can you reproduce the problem on GNU/Linux?
> Make the same change to 'configure', then run
> something like this:
>
> emacs_cv_var_doug_lea_malloc=no ./configure
> MALLOC_PERTURB_=23 make

Ah, didn't know about MALLOC_PERTURB_, thanks. You're reminding me that
I also have strict malloc settings on OpenBSD (ln -s S /etc/malloc.conf).

Anyway, I have followed your instructions and I get a segfault as soon
as the build process fires bootstrap-emacs on Debian Squeeze (amd64).
I have attached the gdb backtrace for both OpenBSD and Debian at the end
of this mail.  Similarly, temacs (appears to) run fine since a few
minutes.

Perhaps should I add that those two tests were launched with:

$ ./configure --without-x --without-dbus --without-gconf \
--without-gsettings && make
$

and

$ emacs_cv_var_doug_lea_malloc=no ./configure --without-x \
--without-dbus --without-gconf --without-gsettings && MALLOC_PERTURB_=23
$

respectively - but I don't think that changes much in the end.

> This tries to set up the same sort of environment, with
> address space randomization and everything.  If this fails in
> a similar way, it might be easier to debug than the OpenBSD
> port would be.  And if it doesn't fail it's an indication that
> the problem might be more on the OpenBSD side than in the Emacs
> side.

OK.  I hadn't thought about trying on other OSes

By the way, to makes things a bit clearer: I want to try to make it work
with the system allocator for no other reason than "I feel it should
work".  Emacs works fine for me as is and you folks probably have other
fish to fry.  But if people have some hints and/or want to help
debugging this, I'd be glad of course.