On the testsuite vs locale models

[Paolo Carlini]

Hi,

yesterday, working on libstdc++/8887 (the fix is really trivial but 
finding it required some gdb work ;) it occurred to me that perhaps I 
could propose something about the 22_locale tests.

Wouldn't be worthy to separate those tests which are supposed to pass 
only with the gnu locale model from the others?

For instance, the issue put forward by libstdc++/8887 is currently 
*already* tested for but since within the generic locale model the whole 
testfile fails for dozen of reasons (there are many named locales tests 
in it), it was not possible to notice it  :-(

Also, I remember a message from Mark, which didn't like those 26 
"expected" ;) XPASSes for the gnu locale model... Perhaps it's time to 
do something about it!

Ideas? Comments?

Ciao,
Paolo.