Re: variable substitution refs [PATCH]

[Tom Tromey]

>>>>> "Richard" == Richard Boulton <address@hidden> writes:

Richard> if COND
Richard> var = foo.c
Richard> else
Richard> var = foo.c
Richard> endif
Richard> foo_SOURCES = $(var:.a=.b)
Richard> sets foo_OBJECTS to empty.

Thanks.  I agree with your analysis and your patch.
However, I'm checking in a slight variant.  I made a new global to
hold the pattern we use to match substitution references.
Doing that is somewhat self-documenting; it means we don't have to add
a comment explaining what some random regular expression is meant to
do.

Richard> I attach a patch which adds a regression test and a fix.
Richard> Unfortunately, the patch depends on my earlier submitted
Richard> patch having been applied already.

Nothing unfortunate about that.  However, I got the messages in
reverse order.  Akim numbers his patches to handle this problem; that
is one convenient approach.

Richard>  cond13.test \
Richard>  cond14.test \
Richard>  cond15.test \
Richard> +cond16.test \

There is no cond15.test.
Odd.

Richard> +val=`$MAKE echorule`;

I did the same GNU-make-ification here.

Tom