This is the mail archive of the
mailing list for the libstdc++ project.
Re: vector _IterCategory undeclared
- To: gcc-patches at gcc dot gnu dot org>,<libstdc++ at gcc dot gnu dot org>
- Subject: Re: vector _IterCategory undeclared
- From: Stephen M. Webb <stephen at bregmasoft dot com>
- Date: Fri, 6 Jul 2001 09:51:33 -0400
- Cc: Gabriel Dos Reis <Gabriel dot Dos-Reis at cmla dot ens-cachan dot fr>
- Organization: CRYPTOCard Corporation
- References: <Pine.BSF.email@example.com>
- Reply-To: stephen at bregmasoft dot com
On Fri, 06 Jul 2001, Gerald Pfeifer wrote:
> The remaining question is: Is the libstdc++ testsuite so weak that this
> was not caught (I really cannot believe it)? Or, did someone not properly
> run the testsuite at some point, thus missing this regression?
I submitted the original patch that introduced the bug. I ran the libstdc++
testsuite and received no [new] errors. I did not run the entire gcc testsuite
since I use my machine for other things over the 20 hour period it takes.
Doing so would have caught the two errors I introduced.
I would submit that the libstdc++ testsuite is woefully inadequate. Because I
have additional patches similar to the first in the works and have been burnt
once, I'm feverishly trying to generate additional testcases that should catch
(indeed, have already caught) problems. Trouble is, I'm spending more time
devising testcases than in actually doing the patch, perhaps as it should be,
but I still don't think I can guarantee complete coverage. There are an
exponential number of ways to instantiate templates, they are only
instantiated if used, and semantic check are only performed on the template
instance, so test coverage is almost necessarily incomplete.
It would be a good recommendadtion, however, to run the entire gcc testsuite
before submitting a patch to libstdc++, since it provides additional testcases
in areas the libstdc++ testsuite does not cover.
Stephen m. Webb