Re: Sparc & Alpha on codecvt_unicode? (was: Re: failure notice)

[Alexandre Oliva <aoliva at redhat dot com>]

On Mar 25, 2002, Paolo Carlini <pcarlini@unitus.it> wrote:

> Alexandre Oliva wrote:
>> On Mar 22, 2002, Paolo Carlini <pcarlini@unitus.it> wrote:

>>> (by the way, I cannot find myself relevant Sparc-linux testresults,
>>> only Alpha)

>> Interesting...  I have no idea of why my sparc-linux-gnu box is not
>> running (or reporting) libstdc++-v3 test results.  I know for sure
>> it's been having hardware problems lately, but this should not lead to
>> consistent failure to check libstdc++.  I'll keep an eye on it while
>> building this week's snapshot.

> Thanks Alexandre. We desperately need results from other big-endian
> gnu-linux platforms besides powerpc and s390 (which we already know
> are ok), since with Ulrich Weigand we are not completely sure to have
> fixed in the correct way 22_locale/codecvt_members_unicode_xx.cc.

Ok, I figured out what the problem is, but it's going to be tricky to
fix.

Basically, the problem is that I don't have a usable dejagnu already
installed on my sparc box.  This shouldn't be a problem because I'm
building uberbaum, so I get tcl, expect and dejagnu all from there.

the problem is that the rule generated by automake to set up
TCL_LIBRARY in this case appears to assume that directories being
tested are built as siblings of expect, but in this case we're
potentially deep in one of the multilibs built for a target library.
Uh oh.

The solution is for me to set TCL_LIBRARY myself when running tests,
but ideally automake should take care of this.  Tom, do you know
whether automake has had any changes in this regard lately?

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer