This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
RE: HPUX10.20 REGRESSION, EGCS-1.0.3A TO 19980803 -Reply
- To: tprince at cat dot e-mail dot com
- Subject: RE: HPUX10.20 REGRESSION, EGCS-1.0.3A TO 19980803 -Reply
- From: Joe Buck <jbuck at synopsys dot com>
- Date: Thu, 6 Aug 98 10:16:10 PDT
- Cc: egcs at cygnus dot com
> I guess your point is that egcs-1.0.3a might have failed to show
> errors, as a side effect of some other failure, the most likely such
> possibility being that that particular test was skipped in 1.0.3a
> due to failure to compile.
Yes. Basically I'm asking that for any apparent regressions, the test
runner should look in gcc.log or g++.log and see what egcs-1.0.3a (or
gcc-2.8.1, for those running test on it) really did when running the
egcs-1.1-snapshot test suite. Here "regression" means either that 1.0.3.a
says PASS and 1.1-snapshot says FAIL, or 1.0.3.a says XPASS and
1.1-snapshot says XFAIL.
It's not necessary to look at the full set of failures, but I found, for
the sparc Solaris 2.5.1 tests, that all the regressions were false. Here
"false" meant that a test said XPASS but the compiler then core dumped, so
it is not true that 1.0.3a handled any tests correctly that
egcs-1.1-snaphsot failed on. So what I'm asking is that if you find
regressions you check to make sure that they are "real".