What could be the problem when make gets into a loop in sax, continuously compiling facreg.cxx?

[Norbert Thiebaud]

On Fri, Mar 2, 2012 at 4:52 AM, Matúš Kukan <matus.kukan at gmail.com> wrote:
> On 2 March 2012 11:31, Tor Lillqvist <tml at iki.fi> wrote:
>> Is there a way to do "make clean" only for the host platform, not the
>> build platform? Is there a way to make clean only for object files,
>> and especially not the stuff generated from IDL files?
>
> Well, I would push:
>
> --- a/Makefile
> +++ b/Makefile
> @@ -300,8 +300,6 @@ clean: clean-host clean-build
>
>  clean-host:
>        rm -fr $(SRCDIR)/*/$(INPATH)
> -       rm -fr $(SRCDIR)/workdir/*
> -       rm -fr $(SRCDIR)/solver/$(INPATH)
>        rm -fr install
>
> I don't know what was Norbert's intention with that.

make clean should remove everything built by make
make dist-clean should do make clean + remove stuff generated by autoconf

sure clean-host: should try to delete only 'host-side' objects.. so
$(SRCDIR)/workdir/$(INPATH)/*

and clean-build
should delete
$(SRCDIR)/workdir/$(INPATH_FOR_BUILD)/*
$(SRCDIR)/solver/$(INPATH_FOR_BUILD)/*

and there is the special case of the delivery of dmake... (although I
personally don't consider dmake as part of the build, but part of
autoconf/setup, so I would not clean it with make clean).

That being said, one could find it useful to have a 'weak-clean'
target to remove only .o and generated .h/.c
in workdir is should be quite easy... in solver (for delivered
generated .h for instance) that could be much more tricky... but since
it would be a 'best-effort' partial solution... that may be good
enough...

On that topic... motivated by windows performance, I'm chewing on the
idea of having an 'incremental build' tinderbox... that would not do a
make clean, unless configure.in/config_host.in has not changed...

There will be hick-up, but then fixing them could lead to a reliable
incremental build that could be generalized to non-windows build...

Norbert