Re: [PATCH]: Remove -DGENERATOR_FILE from rtl.o, bitmap.o targets

[Mohan Embar <gnustuff at thisiscool dot com>]

>> As for the failure, rtl.c includes system.h which includes <sys/wait.h>
>> if HAVS_SYS_WAIT_H is defined. <sys/wait.h> exists on i686-pc-linux-gnu
>> but not on i686-pc-mingw32. However, -DGENERATOR_FILE causes
>> rtl.c to incorrectly include bconfig.h (which defines HAVE_SYS_WAIT_H)
>> instead of config.h (which doesn't), thereby freaking out the build of rtl.o.
>
>This is a believable consequence of the spurious -DGENERATOR_FILE for
>rtl.c, but *not* for the configuration you have described.  You have
>build==host, so bconfig.h and config.h should be identical.  It's
>tconfig.h that should describe mingw32, and rtl.c is never compiled
>for the target.

Argh. You're right. I'm getting my cross builds and native builds messed
up. In other words, the problem probably didn't manifest itself for the
(i686-pc-linux-gnu,i686-pc-linux-gnu,i686-pc-mingw32) build, but probably
when I was using the (i686-pc-linux-gnu,i686-pc-mingw32) cross to build
the (i686-pc-linux-gnu,i686-pc-mingw32,i686-pc-mingw32) compiler. I've
gone through so many builds and different snapshots (due to the current
HEAD breakage for reasons yet unknown) that I lost track of things.

If you agree that this patch is necessary for the
(i686-pc-linux-gnu,i686-pc-mingw32,i686-pc-mingw32) build, let me know.
Otherwise, I think that a cvs -D 2004-09-04 will allow me to builld
all the configurations I mentioned before and I'll confirm for you that
(i686-pc-linux-gnu,i686-pc-linux-gnu,i686-pc-mingw32) works even without
this patch. It's your call. Thanks again for looking at this.

-- Mohan
http://www.thisiscool.com/
http://www.animalsong.org/