Fix all check-target-libstdc++-v3 ICEs

Fri, Jul 04, 2008 at 12:21:15AM +0200, Paolo Carlini wrote:
Jakub Jelinek wrote:
>and except for they all fail
>on the trunk as well (the test have bogus expectations for glibc 2.8+).

Could you possibly clarify a bit this glibc2.8+ thing? I exactly changed
(see libstdc++/36616) the expectations for glibc2.8+ exactly because we
were seeing fails on openSUSE 11.0, which also uses glibc2.8. Which is
the official localedata, the one coming with openSUSE 11.0 or the one
coming with Fedora9? Since you are also working on glibc I would
appreciate an official word about that, then we can just XFAIL for the
rest of the world and be done with it, after it's just a minor testing
issue

I'm very sorry, after checking the facts it is me who is backing the
fr_FR changes out temporarily, because it came too late in the 2.8
cycle and could break a lot of stuff.

Instead of conditionalizing the testcases on glibc version, couldn't
you better just pick up some locale that hasn't changed time LC_TIME
over the time? Say nl_NL, es_ES, etc.?

Jakub

Jakub Jelinek wrote:
I'm very sorry, after checking the facts it is me who is backing the
fr_FR changes out temporarily, because it came too late in the 2.8
cycle and could break a lot of stuff.
>
Instead of conditionalizing the testcases on glibc version, couldn't
you better just pick up some locale that hasn't changed time LC_TIME
over the time? Say nl_NL, es_ES, etc.?

Thanks for the clarification, Jakub.

I agree, changing locale makes sense, I don't think there is anything in
those tests depending critically on the exact locale. I'll see what I
can do

Paolo.