On Mon, Sep 23, 2013 at 03:55:26PM +0200, Marc Glisse wrote:
> On Mon, 23 Sep 2013, Paolo Carlini wrote:
>
> >>There are a lot of targets using unsigned int for size_t, which would
> >>have been uncovered by proper testing.
>
> We can't test all patches on 3-4 different targets... It wasn't
> obvious this patch could be that sensitive to the target.
>
> >That's true, just remember to test *both* -m32 and -m64, for non
> >trivial changes.
>
> So how do you do that in practice ? Is it done by default if
> multilib is enabled? You also mentioned doing something special to
> check debug/profile modes recently, is there a make target to really
> perform all the tests necessary for a submission?
It isn't done by default, but you can easily do that, by running
make check RUNTESTFLAGS='--target_board=unix\{-m32,-m64\}'
(either toplevel, or more specific, e.g. just in libstdc++-v3/
dir, or even
make check-gcc RUNTESTFLAGS='--target_board=unix\{-m32,-m64\} dg.exp=pr12345.c'
etc.).
Or of course you can do two separate bootstraps/make check (that is what I'm
usually
doing, so that both bootstraps are tested).
>
> http://gcc.gnu.org/contribute.html has an outdated section on
> testing. It mentions that you should do a bootstrap for a change to
> the C front-end (should also be for the C++ front-end and I guess
> libstdc++ even if it isn't used much inside gcc).
Jakub
