This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: gcc/gcc/testsuite ChangeLog gcc.dg/20040203-1. ...
On Tue, 3 Feb 2004, Michael Matz wrote:
> On Tue, 3 Feb 2004, Paolo Carlini wrote:
>
> > However, using the PR number is *not* ok, right? There are no other
> > examples in the testsuite.
See gcc.dg/pr9365-1.c, gcc.dg/pr10392-1.c, gcc.dg/pr11864-1.c, various
examples in the Fortran testsuites (using different naming styles, some
don't have the opening "pr").
> Well, such names make much more sense than dates, which bear no
> interesting information at all, and should be avoided for future
> testcases. We also name testcases according to features they test, so
> the most natural names of testcases resulting from bugzilla entries are
> the bug numbers itself (which then also makes it easy to look up that
> bugreport, i.e. history of the test). All IMHO.
Descriptive names (referring to features rather than numbers) are the most
preferred, something like truthvalue-{array,struct,union}-1.c for the main
testcases for the improved error messages. Then I'd prefer PR number
names to date names (some bugs don't have meaningful features being
tested, a PR number is more meaningful than reload-1234.c, a large part of
the function of testcases for optimiser and middle-end bugs is that they
exercise historically fragile parts of the compiler so a new bug causing
them to regress may well be unrelated). Date names make sense for bugs
with no particular feature tested and no PR (I don't see a real need to
open a PR if you discover a bug and fix it immediately).
--
Joseph S. Myers
jsm@polyomino.org.uk