This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: PR 5432: Thread safety patches
- From: Benjamin Kosnik <bkoz at redhat dot com>
- To: Craig Rodrigues <rodrigc at mediaone dot net>
- Cc: libstdc++ at gcc dot gnu dot org
- Date: Tue, 22 Jan 2002 18:40:02 -0800 (PST)
- Subject: Re: PR 5432: Thread safety patches
> Is the patch OK for commit? It it OK for 3.0-branch?
This patch looks pretty good, I think. I'm not quite sure if all the bits
should go in at once (this + the sentry work).
I'm in favor of this testcase getting into the testsuite in some form. A
lot of the good MT testcases are getting lost, and they need to start
being saved in the testsuite directory, even if they are not regularly
run.
Loren and Phil can probably help, but it seems like "benchmark" (or
performance_analysis or whatever) and "threads" subdirectories in the
testsuite directory are inevitable. Or, one "optional" subdirectory with
"threads" and "benchmark" subdirs in it. (Which would keep the current
test harness from running these tests by default.)
I guess benchmark isn't a great name as a lot of these tests aren't
really benchmarks.
I'm unsure how to tell, via dejagnu, if threads are enabled.
In general, this patch is not a great idea for 3.0-branch, I think. Does
changing the data member type change the size of the object? Can this be
said with certainty on all platforms, especially non-x86? Hmm...
-benjamin