This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Recent 4.1 breakage...


Benjamin Kosnik wrote:

But like this, simply commenting out the initial #undef one gets the original, finer grained because of the individual assertions, testcase!

Not quite sure if this is simple enough to actualy do it regularly.

I hope so! I mean, in my mind, regular testsuite runs (at installation time, for instance) should simply check that everything works as expected. Then, while debugging or other special circumstances, the developers can decide to uncomment the #undef and do a finer grained check. For the developers it *is* simple enough ;)

However, as an expedient fix, seems ok. I suppose the goal is to
re-enable this at some later date?


I don't know. The #undef-ed version of the test has the advantage (od course) of being safe, doesn't risk leaking semaphores (the present issue), I think it makes for a good default. Still, the finer grained version (many tests have multiple VERIFY) would be still available to knowledgeable people. I'm happy with that, frankly.

Paolo.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]