David wrote, quoting me:
>> Read and respond to this message at:
>> By: keithmarshall
>> And to add insult to injury, it's even more comprehensively broken,
>> when compiled *natively* with gcc-2.96, on my Mandrake Linux 8.2
>> box :(
> The same with gcc-3.3.5 on Debian, so it's still broken, so to
> speak, up to at least that version. The program _does_ work
> correctly if you compile with the '-std=c99' option. I have
> gcc-2.95 installed, also, and that option is not available for
> that version.
Hmm. Yes, that option *is* available in Mandrake's gcc-2.96, and
it does fix the problem there too :)
> I'm not all that proficient with gcc, but the fact that this option
> was necessary was not immediately intuitive to me from reading the
> round() manpage.
The manpage I have *does* note C99 conformance, but that isn't
sufficient, IMO -- it should also have a `Caveats and Bugs' section,
in which it *explicitly* specifies the requirement to compile with
the `-std=c99' option, (or some other standard conformant option).
In this respect, this is a bad manpage.
The *implementation* is also bad, for without an appropriate `-std'
option, the application will still compile and link without a murmur,
only to fail miserably, and *incomrehensibly*, at run time.
But this isn't the right place to complain of this. Is it a GCC
issue? Or should it be reported somewhere else? I simply don't know
where to go with the bug report :( (And that's another omission from