RE: depcomp problem [Fwd: Trying to compile latest CVS on old SC Ounixware 2]

[Cameron, Steve]

Tom Tromey wrote:
> 
> >>>>> "Steve" == Cameron, Steve <Steve.Cameron@COMPAQ.com> writes:
> 
> Steve> Making all in lib
> Steve>         source='argmatch.c' object='argmatch.o' libtool=no \
> Steve>         depfile='.deps/argmatch.Po' 
> tmpdepfile='.deps/argmatch.TPo' \
> Steve>         depmode=none /bin/sh ../depcomp \
> Steve>         ../compile cc -DHAVE_CONFIG_H -I. -I. -I.. 
> -I../src -I../src      -g
> Steve> -c `test -f argmatch.c || echo './'`argmatch.c
> Steve> compile: no `-o' option seen
> 
> I checked in a patch to the `compile' script which ought to fix this.
> 
> I believe the problem you are seeing comes about like this:
> 
> * AM_PROG_CC_C_O rewrites CC to invoke the compile script
> * The compile script barfs if it doesn't see `-o' (and also if it
>   doesn't see a .c file, which you would hit if you got to linking)
> * However, compile is run (via $(CC)) in just this way when invoked
>   from a pattern rule
> 
> The change I committed addresses this.
> Can you try it out?

yes, with some help from Derek.  I'm just using the automake generated
files which are part of the CVS source now, I don't have automake on that
system. 

> 
> This is all horribly obscure. 

Can't argue with that.

But osn't that kind of the point of automake?  
To protect Derek and other developers from having
to think about my horribly obscure platform? :-)
(Or maybe it does much more, and I'm thinking of autoconf.)

-- steve

> Perhaps it isn't worthwhile for us to
> try to support all the features we support at once.  In this case,
> automatic dependency tracking and parallel `make' with a vendor
> compiler that doesn't understand `-c -o'.  At this point it's probably
> more work to rip it out.  Plus it is sort of amazing that it can work
> at all.
> 
> Tom
>