Re: [PATCH] #include <stdint.h> in tst-writev.c

[Roland McGrath <roland at redhat dot com>]

I don't follow.  Is one of the headers tst-writev.c already uses actually
supposed to define INT32_MAX?  Looking at the list off hand, I don't think
that's so.

If it just happens to be defined implicitly by the glibc -include stuff or
build-time wrapper headers by happenstance, then that is not a reason to
leave out the proper #include from the file itself.  The test files should
be written so they are close to being standalone code that could be built
outside glibc (modulo test-skeleton.c, of course).  People obviously have
occasional need to turn one of these tests into a standalone build for
diagnosing issues, and there's no reason to make life more obfuscated for
them.  Moreover, if the vagaries of implicit header inclusion in the libc
configuration mess change in the future, we shouldn't have to run around
fixing all the sloppy #include usage in individual tests or library routines.


Thanks,
Roland