This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Implement -MF <output-filename>
- To: Neil Booth <neil at daikokuya dot demon dot co dot uk>
- Subject: Re: Implement -MF <output-filename>
- From: "Zack Weinberg" <zackw at Stanford dot EDU>
- Date: Sat, 6 Jan 2001 10:27:54 -0800
- Cc: gcc-patches at gcc dot gnu dot org, Tom Tromey <tromey at cygnus dot com>
- References: <20010106140754.A25638@daikokuya.demon.co.uk>
On Sat, Jan 06, 2001 at 02:07:54PM +0000, Neil Booth wrote:
> Unfortunately this complicates things w.r.t. -MD and -MMD. The "cpp"
> and "gcc" drivers would take these latter options without an argument,
> and pass them to cpp0 with an argument.
>
> This meant cpp0 was providing multiple ways to express the output file
> name (not to mention the vaguely documented environment variable
> OUTPUT_DEPENDENCIES and the undocumented SUNPRO_DEPENDENCIES).
>
> So, I scrapped -MD and -MMD from cpp0 (worthwhile simplification, in
> fact), and re-did the driver specs to cause -MD and -MMD send "-M -MF"
> or "-MM -MF" instead. This seems cleaner and more logical. It also
> eliminates much of the confusion that existed with -MG and -MD / -MMD.
>
> The command line -MF overrides environment variable designations,
> since this seems to be standard practice.
>
> Zack, do you think I'm going the right way with this? I'll wait for
> your confirmation before committing.
I like it overall.
It is difficult for me to comprehend your patch since it's got
substantial code rearrangement on top of the semantic changes.
There's no change log for gcc.c.
-MD expands to -M -MF %b.d. %b is the basename of the input file -
no directory component. Is this what we want? We just went through
changing the target name in dependency output to include the
directory. However, since -MD's always done this, we can worry about
it later.
Do we have documentation in cpp.texi for all these new options?
zw