[OE-core] [PATCHv3] man: fix several annoying compile/build warnings

[Richard Purdie]

On Wed, 2016-03-30 at 08:32 +0100, Richard Purdie wrote:
> On Tue, 2016-03-29 at 15:27 -0700, Bill Randle wrote:
> > Fixed the build error when building man.config.5 (a remnant of a
> > long
> > ago previous patch). Optimized the manpages Makefile for parallel
> > builds. Drop a FHS patch that is no longer needed, as the standard
> > Makefile puts the man pages in the proper location. Also, fix
> > compile
> > warnings in a couple other files.
> > 
> > [YOCTO #9341]
> > 
> > Signed-off-by: Bill Randle <william.c.randle at intel.com>
> > ---
> > [Note: should be applied after previous man patch "fix src/Makefile
> > to work with parallel make".]
> > 
> > v3: clean up patch submission for mailing list
> 
> These two were applied in the build yet we saw:
> 
> https://autobuilder.yoctoproject.org/main/builders/nightly-x86-lsb/bu
> il
> ds/718/steps/BuildImages/logs/stdio

The man failure worried me a bit so I had a look at the code. The
configure script looks good with its BUILD_CC and CC, however it goes
on to build things with $CC then run them which in a cross environment,
isn't going to work.

The reason we only saw failures in qemux86-64 with poky-lsb is because
that is the only case where the binaries generated with $CC could
actually run (in the normal poky case, the interpretor isn't in /lib64/
but /lib/).

On my local builds in conf_script I see:

s, at DEFS@, -DUSG -Duid_t=__uid_t -Dgid_t=__uid_t -DALLOCA_MISSING -DNONLS -DNOGETOPT -DDO_COMPRESS,

and on the failed builds on the AB:

s, at DEFS@, -DSTDC_HEADERS -DTERMIOS_HEADER -DPOSIX -DDO_COMPRESS,

with the -DNONLS being necessary to avoid build failures.

I suspect we're going to have to seriously patch the configure script
to remove several of the tests and then set the options to the correct
things manually.

I haven't looked further at what other nastiness may lurk beyond this
but at least we have a clear understanding of why its breaking now...

Cheers,

Richard