This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH v3 8/9] aarch64: Consolidate NPTL/non versions of clone
- From: Richard Henderson <rth at twiddle dot net>
- To: Andreas Schwab <schwab at linux-m68k dot org>, Marcus Shawcroft <marcus dot shawcroft at gmail dot com>
- Cc: GNU C Library <libc-alpha at sourceware dot org>, Richard Henderson <rth at redhat dot com>
- Date: Thu, 29 May 2014 08:34:44 -0700
- Subject: Re: [PATCH v3 8/9] aarch64: Consolidate NPTL/non versions of clone
- Authentication-results: sourceware.org; auth=none
- References: <1401046909-25821-1-git-send-email-rth at twiddle dot net> <1401046909-25821-9-git-send-email-rth at twiddle dot net> <CAFqB+Pw0ae1k7EWE4PnQcLzio86mHHdkBsTjA-CMTbJ86i+vwg at mail dot gmail dot com> <m2lhtlc85z dot fsf at linux-m68k dot org>
On 05/29/2014 01:15 AM, Andreas Schwab wrote:
> Marcus Shawcroft <marcus.shawcroft@gmail.com> writes:
>
>> I notice that some of the targets make thread_start local (x86_64,
>> i386, hppa, tile, mips) while others do not (alpha, m68k, s390).
>
> An assembler label is always local by default.
Indeed, but what Marcus meant was "really" local, i.e. ".L" or "1:"
type symbols, vs "static" symbols that are local in the elf sense but preserved
by the assembler.
My guess is that with the pre-dwarf alpha unwind info one had to use a
different ".ent" directive, which implied that it was impossible to put the
unwind info for thread_start into the same function as clone. And then others
copied how Alpha had set it up.
Personally, I like having "thread_start" at the top of the backtrace. But I
can see how others might prefer "clone", as some indication of how that happened.
r~