This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc 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] |
On Mon, Oct 06, 2014 at 09:33:33AM +0100, Will Newton wrote: > On 3 October 2014 20:36, Roland McGrath <roland@hack.frob.com> wrote: > > The dup2 hack is either cute or horrible (likely both). I don't think we > > should do anything like that in one random test. Rather, we should settle > > on a uniform policy for tests and using assert. It could be that we never > > use assert. Or it could be that we use assert freely and just make > > test-skeleton.c always do the dup2 hack itself. Or it could be something > > else I haven't thought of off hand. But let's get a plan. > > One option could be to use a hand-rolled assert function which is what > most unit testing libraries do. This would also give scope for adding > more esoteric asserts that can output more detail about the failure. Agreed, but to be clear, the hand-rolled assert should always exit normally, i.e. not with an abort. An abort precludes any cleanups that may be necessary for a test. In fact asserts that result in such abnormal termination should be actively discouraged. On whether we should redirect stderr to stdout, I am not very excited about it. I don't want to exclude the possibility of wanting to use stderr some time in future for the tests. Siddhesh
Attachment:
pgpCvoVlWm0LS.pgp
Description: PGP signature
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |