Re: please test libio

[Roland McGrath]

> Basically, I have done a native compile (over NFS) of a libio-enabled
> glibc.  After a make check,
> 
> find . -name \*out -exec wc '{}' \;
> 
> tells me that all the test result files except for a few are 0 bytes.

Many of the test out files are normally empty, but many others are not.
However, I think it's the case that they should all exit abnormally if they
fail, so `make check' should obviously fail if something is wrong, in theory.

The best answer is that you should do a native Linux build of glibc and
compare the `make check' results from that to what you get on the Hurd.
But I can tell you now that if e.g. all the stdio-common/ and libio/
test output files are empty, then there is definitely a bug.

> I notice that all of the tests seem to use the newly built ELF loader,
> and the apps are all generated using the newly built libc.  Doesn't
> that mean that make check shouldn't run?

I'm not really following your question here.  When libc runs a program
built against the newly-built libc, the only way it can do that without
installing first is to run the new ld.so directly.

> I still have all the build logs, and test logs, etc if you want me to
> send them to you.

Never hurts.