This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Failure to dlopen libgomp due to static TLS data
- From: Andrew Haley <aph at redhat dot com>
- To: Rich Felker <dalias at libc dot org>
- Cc: Jakub Jelinek <jakub at redhat dot com>, Ulrich Weigand <uweigand at de dot ibm dot com>, libc-alpha at sourceware dot org, gcc at gcc dot gnu dot org, rth at redhat dot com
- Date: Sat, 14 Feb 2015 13:46:40 +0000
- Subject: Re: Failure to dlopen libgomp due to static TLS data
- Authentication-results: sourceware.org; auth=none
- References: <201502121519 dot t1CFJMAe018776 at d03av02 dot boulder dot ibm dot com> <20150212160959 dot GS23507 at brightrain dot aerifal dot cx> <20150212161145 dot GD1746 at tucnak dot redhat dot com> <20150212161617 dot GU23507 at brightrain dot aerifal dot cx> <54DCEF90 dot 6090700 at redhat dot com> <20150212232756 dot GZ23507 at brightrain dot aerifal dot cx> <54DDC009 dot 3030505 at redhat dot com> <20150213222357 dot GH23507 at brightrain dot aerifal dot cx>
On 02/13/2015 10:23 PM, Rich Felker wrote:
> It seems misguided to try to save 0.5% of a 500ms startup time by
> choosing a hackish TLS model
No. I don't want a hackish TLS model. I want the bug fixed so that
the correct dynamic TLS model is fast if there is static TLS space
available, and automagically falls back to slow dynamic TLS if it's
not. That's how it worked for me last year, and lxo's patch is
supposed to fix it.
Andrew.