On Sat, 27 May 2000, Ralf Baechle wrote:
> On Fri, May 26, 2000 at 11:32:06AM -0700, Ulf Carlsson wrote:
>
> > > Can anyone please suggest a workaround? I'm stuck right now.
> >
> > I think there should be a line that says
> >
> > CFLAGS_r4k_genex.o := -P
> >
> > in the Makefile for the kernel directory. If you do that cpp doesn't output
> > line numbers.
>
> The workaround for the affected files is already in the CVS Makefile
> since a week. I hope Ulf soon comes with a fix for the problem in
> binutils ...
I haven't found a nice solution directly in ecoff.c, but you can easily get
around the problem by not generating line number debug information when you're
expanding macros. I don't think this has any unpleasant side-effects for
other architectures either. You basically don't care about the line number
information cpp inserts into macros anyway. I could compile some new RPM if
you really want to try this...
It is a design problem with the ecoff code and not just a silly bug that I
could fix without affecting anything else.
Ulf
Index: read.c
===================================================================
RCS file: /cvs/src/src/gas/read.c,v
retrieving revision 1.19
diff -u -p -r1.19 read.c
--- read.c 2000/04/21 20:22:22 1.19
+++ read.c 2000/05/30 09:43:20
@@ -875,8 +875,6 @@ read_a_source_file (name)
c = *input_line_pointer;
*input_line_pointer = '\0';
- generate_lineno_debug ();
-
if (macro_defined)
{
sb out;
@@ -899,6 +897,8 @@ read_a_source_file (name)
continue;
}
}
+
+ generate_lineno_debug ();
if (mri_pending_align)
{
|