Enhancement patch for cpp

[Dave Brolley]

Ben,

Looks like the consensus is that this warning needs some restrictions in
order to prevent it from becoming annoying. I think that Zack's suggestion
to control it via a -W option is a good one. I think it should also be
enabled by -Wall.

Dave

Zack Weinberg wrote:

> On Mon, 18 Jan 1999 10:31:07 -0500, Dave Brolley wrote:
> >Zack Weinberg wrote:
> >
> >> >> 1999-01-15  Ben Elliston  <bje@cygnus.com>
> >> >>
> >> >>         * cppfiles.c (append_include_chain): Emit a warning if the
> >> >>         directory specified in a -I flag does not exist.
> >> >>
> >> >>         * cccp.c (new_include_prefix): Likewise.
> >>
> >> I'm not sure about this patch.  Ignoring nonexistent directories is
> >> done because the default include path has several dirs on it that
> >> often don't exist, such as /usr/local/include and
> >> /usr/cpu-company-system/include.  They are processed through the same
> >> function, and warning about them would be confusing.
> >
> >Well, it's not too late....I haven't installed the patch yet. Zack's
> >point is a good one. Perhaps the warning should be limited to paths
> >specified using compile time options.
>
> That might make sense.  Now, is it really useful to have this warning?
> It's easy to check if directories exist.  On the other hand, the
> warning might uncover some mistaken expectations.  Maybe it should be
> controlled by a separate -W switch.
>
> I'd appreciate it if you would hold off any cpplib changes for the
> moment, I have a bunch of stuff stacked up and the first piece is a
> major code rearrangement (which will be mailed later today, after the
> tests finish).
>
> zw