Re: building with ccache w/gcc4.8

[Jonathan Perkin <jperkin%joyent.com@localhost>]

* On 2013-08-14 at 14:06 BST, Richard PALO wrote:

> Le 14/08/13 14:11, Jonathan Perkin a écrit :
> 
> >>In the meanwhile I noticed in mk/transform-gcc that both
> >>`-specs filename` and `-specs=filename` are passed through.
> >>
> >>I'm curious as whether the former should now invoke `transform_fail`
> >>instead of passing it in?  Are there any real world cases where this
> >>is necessary?
> >
> >We definitely shouldn't fail, '-specs file' is perfectly legitimate
> >GCC syntax.  What we could do is transform '-specs ' to '-specs='.
> >
> >However, the real bug is in ccache, which should be fixed to correctly
> >handle the '-specs file' case.
> >
> There is also transform_pass_with_warning...
> But that seems to only put a warning into .work.log
> 
> But this is exactly the reason I ask, in other words, I cannot seem
> to find anything but specifically '-specs=' as the correct syntax,
> in particular as found in the "Using the GNU Compiler Collection"

No, it's a getopt_long implementation detail.

> Do you happen to have a reference of something with '-specs
> filename' other than these aforementioned pkgsrc buildlink3.mk
> files?

I think for now let's just change the pkgsrc invocations to '-specs='
and then if we hit any problems in specific packages in the future we
can take it from there.

-- 
Jonathan Perkin  -  Joyent, Inc.  -  www.joyent.com