This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Don't set function's DECL_NAME to an EXPR_WFL
Alexandre Petit-Bianco <apbianco@cygnus.com> writes:
> That's because it hasn't been substitued with the DECL_NAME before
> going through expand_expr.
It is still wrong to set the DECL_NAME to an EXPR_WFL.
It breaks debug_tree. It is fragile. It is a kludge.
I suggest not wasting time kludging around something we know is
wrong, unless it really is needed.
> I can fix that if you provide me with a test case.
Well, you can try compiling
http://www.bothner.com/~per/test/kawa-1.6.71-compiled.zip
However, it does seem like wasted time. We know that this patch
is needed sooner or later. So if there are negative consequences
from it, we will need to fix them sooner or later, so doing do
now is not wasted time. Also, if there are negative consequences,
they don't seem to be critical - e.g. libgcj compiles ok. On the
other hand trying to find some alternative work-around for the patch
is likely to be unproductive - fixing the symptom and not the cause.
> We relie on it for precise error message. The front-end parser relies
> on it a lot.
Where? Then the parser shold be fixed.
Having column numbers in error messages is nice, but it is not
necessary. We should phase out the existing kludges and replace
them with something clean.
> Removing it seems to be too risky for 3.0.
>
> I'd prefer you hold this patch back.
Well, if I have time, I'll try to see what happens with this
patch reverted. (I'm recompiling now.)
--
--Per Bothner
per@bothner.com http://www.bothner.com/~per/