Re: [Libunwind-devel] Re: [patch 2/2] Allow caller to block signals.

[Paul Pluzhnikov]

On Fri, Sep 25, 2009 at 11:29 AM, Mark Rabkin <address@hidden> wrote:

> Sorry to jump in late here but could you give a couple more sentences of
> clarification of what this patch intends to accomplish?

Which "this" patch? (There are several in this thread.)

> Is this simply a performance gain by allowing the users of
> google::GetStackTrace() to block signals once (and maybe lock a mutex) at a
> higher level of libunwind?

This patch does exactly above:
http://lists.nongnu.org/archive/html/libunwind-devel/2009-09/msg00001.html

> Or does this help ensure correctness and more
> complete stack traces?

This one helps correctness (completeness) of the stack traces:
http://lists.nongnu.org/archive/html/libunwind-devel/2009-09/msg00004.html

> I'm wondering if we should turn this option on in our multithreaded servers
> that take a lot of traces (sampling some various activities).

You are likely to want the second patch.
You may or may not want the first patch, depending on whether additional
sigprocmasks matter to you.

Cheers,
-- 
Paul Pluzhnikov