This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Policy for wrong-code bugs
- From: Volker Reichelt <reichelt at igpm dot rwth-aachen dot de>
- To: gcc at gcc dot gnu dot org
- Date: Mon, 12 Jul 2004 22:10:02 +0200 (CEST)
- Subject: Policy for wrong-code bugs
Hi,
our current policy with bug reports is:
Close a PR, if the bug
a) is not a regression, and
b) is fixed on mainline.
IMHO this is not a good idea in the case of wrong-code bugs:
I think we should close them, only if they are fixed on mainline *and*
all open release branches (or if explicitly marked as WON'T FIX on those
branches).
Rationale:
An ICE or a rejects-valid is a compile time problem. The user can usually
work around those problems. The most important point is: She/he is aware
of the compiler problem. An accepts-invalid usually does what the user
intended, and in the end it's also the users fault to write invalid code.
But wrong code is different: It's not the users fault, and she/he is not
even aware of a problem. Silently generated wrong code can cause security
problems, endless debugging sessions with heisenbugs etc. In short:
It's really a pain for the users.
In addition, it's a long way from mainline to a stable branch, so users
only have the choice between broken compilers. :-(
Btw, the bug http://gcc.gnu.org/PR16471 made me write this mail
(see also the thread in http://gcc.gnu.org/ml/gcc/2004-07/msg00531.html ).
I'd really like to see this one fixed.
Regards,
Volker