On 05/27/2010 04:08 PM, Todd Rinaldo wrote:
>
> On May 27, 2010, at 4:43 PM, Eric Blake wrote:
>
>> On 05/27/2010 01:23 PM, Todd Rinaldo wrote:
>>> Today, I ran into an issue compiling m4 on RHEL6 beta. I had to apply this patch to make it work or I got errors about S_ISDIR not being defined during linking. Any ideas why this was an issue?
>>
>> Thanks for the report.  Most likely, this is an actual bug in m4 relying
>> on the laziness of older systems that implicitly included <sys/stat.h>
>> inside other headers; contrasted with recent glibc which has
>> intentionally been tightening down system headers to avoid such
>> namespace pollution.  POSIX states that <fcntl.h> may, but not must,
>> include <sys/stat.h>; and it looks like m4.h had another include path
>> that included <fcntl.h> but not explicitly <sys/stat.h>.
>
> Eric, I'm dying to know... How the heck did RHEL 6 m4 get built against RHEL 6 libc when there's no patch in the spec file?

I'm not in charge of the RHEL m4 build, but my guess is that m4 was
built prior to the glibc update that changed the header that would have
broken m4's rebuild.

--
Eric Blake   eblake@redh...    +1-801-349-2682
Libvirt virtualization library http://libvirt.org