Eric Blake <ebl...@redhat.com> writes: > On 05/22/2012 06:21 AM, Simon Josefsson wrote: >> Bruno Haible <br...@clisp.org> writes: >> >>> Hi Simon, >>> >>>> I have reverted the strdup and select patch, and I'll treat >>>> this as a user error. >>> >>> It was a user error. But the select.c patch was good: it fixed an >>> earlier-undetected mistake (that would have likely been a compilation >>> error on Interix). >> >> However Interix probably made NULL available through the sys/select.h >> header so it worked, assuming someone tested the Interix code. > > It may have worked by accident for now, but POSIX doesn't require it to > work, so we are better off using an explicit include that POSIX does > guarantee so that if future Interix ever fixes their headers for POSIX > compliance, we continue to compile. I say we keep the added #include in > select.c.
That is convincing and I agree, so I re-reverted the patch. The same argument doesn't hold for strdup though: strdup is POSIX and it should be available if the user doesn't disable it. So let's leave that patch reverted, I guess? /Simon
