This is the mail archive of the
mailing list for the libc-ports project.
- From: "Will Newton" <will dot newton at gmail dot com>
- To: "Daniel Jacobowitz" <drow at false dot org>
- Cc: libc-ports at sourceware dot org
- Date: Mon, 8 May 2006 15:01:52 +0100
- Subject: Re: FLOATING_STACKS
- Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=h1FjQ/08uMMmotXu1h8xa2ppTJO8smhKo6dvQKYu5zrZcOz1GTcg9i7fcHO4yYNf8MVNXM0ohJmke27RBfaO8fAgFHFQ7YTovapUp8EcQ8/zBmlwaTLOB3YfKGtO4SF5xzHmWfK+dEbrUSeLdHZHq/0OXaCyD4uGO7kGXWz6pj0=
- References: <email@example.com> <20060508132825.GA3589@nevyn.them.org>
On 5/8/06, Daniel Jacobowitz <firstname.lastname@example.org> wrote:
It has very little platform-independent meaning. Basically, if you
have to search the list of stacks to figure out where a thread is, you
shouldn't define it. And it must be constant at compile time. If you
can define THREAD_SELF, then you should also define FLOATING_STACKS.
OK, thanks for the information. I'm far from an expert on LinuxThreads
internals so I'm rather feeling my way with this implementation - I
initially had defined FLOATING_STACKS but not THREAD_SELF. This lead
to rapid death, which I think is to be expected.
For instance, if there's a dedicated register in which you can record
the thread pointer, then you can use this.
What are the ideal requirements of such a register? Reserved in the
ABI? Our ABI is still tentatively open to modification so this may be
a future possibility.
I really don't recommend writing a new LinuxThreads port at this late
date, though. NPTL is a bit harder to port, but well worthwhile.
Unfortunately we are focussing on Linux 2.4 for the near future so
NPTL would appear to be a non-starter.