RE: Patches to C++ front end have caused numerous regressions

["S. Bosscher" <S dot Bosscher at student dot tudelft dot nl>]

Hi,

Mark, you say "numerous" and make it sound real bad, but in your example you
show all _seven_ regressions, compared to test suite runs from two days ago.

OK, so they are regressions and that's not good. But can you please _not_
just revert all that work, especially since you know that the people who
caused these _will_ be very willing to fix them ASAP?

One of the regressions you mention did not show up in my test run
(http://gcc.gnu.org/ml/gcc-testresults/2003-09/msg00440.html) this morning;
this one is new:

  FAIL: g++.old-deja/g++.pt/explicit34.C (test for excess errors)

If you're seeing this one now, then it was introduced _very_ recently, so
that one should be easy to find.

I compared my run to a run by Diego from two days ago
(http://gcc.gnu.org/ml/gcc-testresults/2003-09/msg00356.html), and I see six
new regressions:

FAIL: g++.old-deja/g++.jason/template28.C (test for excess errors)
FAIL: g++.old-deja/g++.other/ctor1.C (test for excess errors)
FAIL: g++.old-deja/g++.other/delete3.C (test for excess errors)
FAIL: g++.old-deja/g++.other/main1.C (test for excess errors)
FAIL: g++.old-deja/g++.pt/memtemp17.C (test for excess errors)
FAIL: g++.old-deja/g++.pt/memtemp7.C (test for excess errors)

IMVHO it looks like it'd be much easier for everyone to just hunt down these
problems and fix 'em...

Gr.
Steven


P.S.
I would like to remind you that even you can't just revert this under the
48hrs rule, since: "_two_ people with write privileges to the affected area
of the compiler determine that the best course of action is to revert the
patch;"






-----Original Message-----
From: Mark Mitchell
To: jh@suse.cz; edelsohn@gnu.org; rth@redhat.com
Cc: gcc@gcc.gnu.org
Sent: 12-9-03 22:27
Subject: Patches to C++ front end have caused numerous regressions


I'm seeing numerous regressions in the C++ testsuite on
i686-pc-linux-gnu.  For example:

  FAIL: g++.old-deja/g++.jason/template28.C (test for excess errors)
  FAIL: g++.old-deja/g++.other/ctor1.C (test for excess errors)
  FAIL: g++.old-deja/g++.other/delete3.C (test for excess errors)
  FAIL: g++.old-deja/g++.other/main1.C (test for excess errors)
  FAIL: g++.old-deja/g++.pt/explicit34.C (test for excess errors)
  FAIL: g++.old-deja/g++.pt/memtemp17.C (test for excess errors)
  FAIL: g++.old-deja/g++.pt/memtemp7.C (test for excess errors)

The problem seem to include missing constructors and destructors, and
internal errors in cgraph_finalize_function.

I will revert the patches causing these problems under the 48-hour
rule if the problems are not corrected by then.

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com