revised patch for command-line discrepancies in cpp driver
Dave Brolley
brolley@cygnus.com
Mon May 31 20:56:00 GMT 1999
I've commited this patch......
Dave
Zack Weinberg wrote:
> On Wed, 5 May 1999 15:04:11 +0200, Marc Espie wrote:
> >On Tue, May 04, 1999 at 06:42:33PM -0400, Zack Weinberg wrote:
> >> >Also, I need to re-check if your patch fixes that, but with the default
> >> >sources, cpp name.s does not work, whereas cpp name.c does.
> >
> >> This is intended behavior of the gcc driver, which is now being used
> >> for cpp as well. name.s is interpreted as `assembly not requiring
> >> preprocessing' just as name.i is `C not requiring preprocessing'.
> >
> >> I can make a case that when the user calls cpp by name, they intend
> >> for the preprocessor to be run anyway. I will leave it to the overall
> >> maintainers to decide which behavior is preferred.
> >
> >> Your workaround is to add `-xassembler-with-cpp' to xf86.rules, or to
> >> rename all the name.s files to name.S.
> >This is highly surprising for a cpp driver.
> >I can see how this makes sense in a weird sort of way.
> >This is a large difference with cccp, this must be documented.
>
> I thought about it some more and decided that the preprocessor should
> always be run. The appended patch is a revision which should achieve
> this. Apply instead of the previous one.
>
> Oh, if and when you have time, I'd like to know whether egcs with
> --enable-cpplib has problems doing a make world on OpenBSD. (This
> should be completely orthogonal to the cpp driver issue.)
>
> zw
More information about the Gcc-patches
mailing list