Re: [3.4 patch?] -save-temps still produces different .s and .o

[Per Bothner <per at bothner dot com>]

Alexandre Oliva wrote:


> Bootstrapping (almost done) and regression-testing mainline on
> i686-pc-linux-gnu.  Ok to install if it passes?
>
> How about the 3.4 branch?  We should really do something about the 3.4
> branch ASAP: either fix -fworking-directory with something like the
> patch below, or take it out.  It would be a shame to release 3.4 with
> a new feature already broken.  The only change that doesn't apply
> there is the one that removes const from cpp_get_line_maps, since the
> branch doesn't have the const there.

I really don't like this change.  Especially for 3.4 changing the
parse_file call-back would be undesirable.

Instead, I think we should change toplev.c so lang_dependent_init
is called after process_options (and specifically the post_options
hook).  The version of the patch I did for the compile-server
does this.  I'll try to for the 3.4 or mainline version.
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/