This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Testsuite policy - do we care if obscure options work?
- To: David dot Billinghurst at riotinto dot com dot au, mark at codesourcery dot com, tprince at computer dot org
- Subject: Re: Testsuite policy - do we care if obscure options work?
- From: Mike Stump <mrs at windriver dot com>
- Date: Thu, 15 Jun 2000 18:48:20 -0700 (PDT)
- Cc: gcc at gcc dot gnu dot org
To answer the subject line, yes, we care that obscure options work (in
general), though they may be of low enough priority that no one fixes
them for a while.
> From: "Tim Prince" <tprince@computer.org>
> Date: Thu, 15 Jun 2000 17:46:20 -0700
> Another question about policy: which obscure combinations of options
> would even be considered for testsuite?
100% of them. Let me clarify. A regression suit should add a
testcase for every bug that has been fixed. For example, in the C++
testsuite, we would have:
// Special g++ Options: -fpic -fno-automatic -O2 -Wall -mregparms
if that was the minimal set of options required to cause the fault.
This means that every combination that has ever failed in some way
will be tested (in the way it failed).
Now, the C testsuite has more than a regression suite in it, and no, I
don't think we should worry about coverage testing with obscure
options on the entire testsuite (c-torture).