In article <[EMAIL PROTECTED]>,
[EMAIL PROTECTED] (Michael G Schwern) wrote:
> On Wed, Dec 15, 2004 at 02:13:38PM -0500, Andrew Dougherty wrote:
> > > On Mon, Dec 13, 2004 at 11:34:05AM +0100, Rafael Garcia-Suarez wrote:
> > > > Only guessing :
> > > > Maybe this warning comes from ExtUtils::MM_Unix::max_exec_len().
> > >
> > > Could be. I wonder why its leaking out of the two evals though.
> >
> > The following patch at least silences it:
>
> I am not enamored of this solution.
>
> If POSIX was loaded and works, it should be quiet about it.
> If it was loaded but doesn't work (ie. its autoloaded functions are non
> functional) then the require should have failed.
>
> I'm still digging into exactly what's happening.
FWIW, this warning does not occur on VMS, most likely because there is
an MM_VMS::max_exec_len() that does not use POSIX. The build log (see
snippet below) does show that autosplitting happens during the running
of POSIX's Makefile.PL, presumably from pm_to_blib. I do not
understand the overall sequence of MM magic well enough to know if the
first reference to POSIX can be deferred until after the autosplit
happens.
Making POSIX (dynamic)
Note (probably harmless): No library found for posix
Note (probably harmless): No library found for cposix
Writing Descrip.MMS for POSIX
cp posix.pm [--.lib]posix.pm
AutoSplitting ../../lib/posix.pm ([--.lib.auto.POSIX])
cp posix.pod [--.lib]posix.pod
. . .
