HEADS DOWN

[Ivan Voras]

Alfred Perlstein wrote:

> Let's clear the issue here...  FreeBSD did not have a choice in the
> matter.

Actually it did. The SIGABORT codepath was only executed in case the
process is running under uid=0, in other cases it would just print out
the warning. I think behaviour such as this is a violation of POLA as
the behaviour is different depending on which users runs it (and if the
allocator can clearly handle the situation for nonprivileged users, it
can also handle it for root, but instead chose to be annoying).

> We can't be "bug for bug tolerant" with Linnex without copying
> their allocator.  

No, but we can with the documented parts, and
malloc-inside-signal-handler support is documented
(http://www.cs.utah.edu/dept/old/texinfo/glibc-manual-0.02/library_toc.html#SEC357).


Even if it's forbidden by POSIX or other standards, FreeBSD isn't
popular enough to be one of the systems that don't support it.

(This being said, since I'm not going to create a malloc implementation
any time soon, it of course depends on the particular author what will
or will not be supported :) ).

> And since our allocator is vastly superior

phkmalloc isn't any more and it remains to be seen for jemalloc :)


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 250 bytes
Desc: OpenPGP digital signature
Url : http://lists.freebsd.org/pipermail/freebsd-arch/attachments/20070513/6e281792/signature.pgp