Re: Unreviewed fixes

[Jan Hubicka <hubicka at ucw dot cz>]

> Jan Hubicka <jh@suse.cz> writes:
> 
> >> Jan Hubicka <jh@suse.cz> writes:
> >> 
> >> > Hi,
> >> > here are few fixes waiting for review I would love to see in mainline :)
> >> >
> >> > Fix for make profilebootstrap failure
> >> >   http://gcc.gnu.org/ml/gcc-patches/2003-09/msg02061.html
> >> 
> >> I do not understand why this fixes the problem.  It looks like you
> >> just rearranged the makefile variables in a way that ought to have
> >> no effect.  Could you explain why this fixes it?
> >
> > It is important to not pass -Werror in stagefeedback as we produce
> > warning about missing profile for each file that contains no function
> > linked into final executable and we compile few of these depending on
> > configuration.
> 
> That wasn't what I was asking.  Perhaps I am being exceptionally dense
> today, but I do not see how your patch causes -Werror not to be
> applied to stagefeedback.  As I said, it looks like it rearranges a
> few makefile variables, with no net effect.

Originally the -Werror got into STAGE2_FLAGS, while I rearange
it to go throught STAGE*_FLAGS_TO_PASS and skip it in
STAGEFEEDBACK_FLAGS_TO_PASS.

But perhaps I can avoid the warning by simply supresing the file to be
output until first function body is produced.  In the case just have
compilation units with no functions in them, it would silence it.  In
the case we have compilation units that may not be linked into final
executable depending on the configuration, I think avoiding -Werror is
only sane way.

I don't want to silence the warning completely as it is common user
error to forget to produce profile data or move them somewhere where we
don't find them.

Honza
> 
> zw