This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: gcc-2.95.3 April 12 incomplete C++ thunks patch breaks Linux/ia32
- To: Franz dot Sirl-kernel at lauterbach dot com, martin at loewis dot home dot cs dot tu-berlin dot de
- Subject: Re: gcc-2.95.3 April 12 incomplete C++ thunks patch breaks Linux/ia32
- From: Mike Stump <mrs at windriver dot com>
- Date: Fri, 19 May 2000 06:23:53 -0700 (PDT)
- Cc: gcc at gcc dot gnu dot org
> Date: Fri, 19 May 2000 12:00:07 +0200
> To: "Martin v. Loewis" <martin@loewis.home.cs.tu-berlin.de>
> From: Franz Sirl <Franz.Sirl-kernel@lauterbach.com>
> Cc: gcc-bugs@gcc.gnu.org
> Frankly, I don't understand at all why the C++ testsuite doesn't at least
> loop over -O0 and -O2 by default, not to speak of the hiding of ICE's in
> XFAIL's...
For a regression suite, I don't prefer adding -O2. For a testing
suite, I prefer adding -O2. Then the question becomes, what is the
g++ testsuite? g++.mike is a regression suite. Most of the g++ is
managed as a regression suite. g++.law, as I recall is more of a
test suite.
Maybe everything that has failed in the past 3 years is a regression
testcase, and everything that has passed in that time is a test suite.
Maybe we don't care anymore about just a regression suite?
And what should make check do? Test, or regression?
Anyway, I thought I'd chime in that around here we always test with a
second pass with -O2 in the g++ testsuite. CPUs are cheap enough
around here. I can dig up the changes if people are interested, let
me know.