Re: Trouble with gnulib/float.h on Intrepid

[Rhys Ulerich]

Hi Eric,

An update...

> $ echo '#include <fcntl.h>' > foo.c
> $ bgxlc_r -E foo.c > fcntl.out

Apparently my user hacked around an issue here.  As produced, the
configure script identified a fcntl.h under
/bgsys/drivers/V1R4M2_200_2010-100508P/ppc/gnu-linux/powerpc-bgp-linux/sys-include/bits/
which broke his compilation due to the header beginning with
#ifndef _FCNTL_H
# error "Never use <bits/fcntl.h> directly; include <fcntl.h> instead."
#endif
That file is attached as fcntl_error.h.

The user mucked around a bit and made the code include a different
fcntl.h from 
"/bgsys/drivers/V1R4M2_200_2010-100508P/ppc/gnu-linux/powerpc-bgp-linux/sys-include/"
which works okay.  It is attached as fcntl_okay.h.
It worked okay.

> $ echo '#include <float.h>' > foo.c
> $ bgxlc_r -E foo.c > float.out

This produces a zero-length float.out.  My user reports there's no
float.h file sitting alongside other headers in the usual places.

> We already special-case AIX, using a different mode of
> '$CC -C -E' to make the preprocessed output more verbose; is there some
> compiler switch for bgxlc_r (maybe -C, maybe some other spelling) that
> makes the output more verbose?

Trying again with -C (documented in the attached man page, in case
that's helpful at all) a la

> $ echo '#include <float.h>' > foo.c
> $ bgxlc_r -C -E foo.c > float.out

produces the output attached as float_with_-C.out.

I'm happy to keep trying to provide what's necessary to get this
sorted, but feel free to bail on the effort whenever this turns into a
time suck.

That said, thank you,
Rhys

fcntl.out
Description: Binary data

fcntl_error.h
Description: Text Data

fcntl_okay.h
Description: Text Data

bgxlc_r.man.gz
Description: GNU Zip compressed data

float_with_-C.out
Description: Binary data