This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [libffi] Patch 14922: fix libgcj EH breakage
On Apr 19, 2004, at 5:02 AM, Andreas Tobler wrote:
Matt Austern wrote:
These sound scary, but they aren't. I have verified that all of
these tests except "indirect" failed before my linkonce checkin, i.e.
that they aren't something I broke. And looking at the indirect.java
source, I just can't believe that that failure can have anything to
do with exceptions or libffi.
Is this the whole patch?
Does NOT work here on darwin-7.3 todays cvs. Unwindtest fails.
Odd. That is not the behavior I see. It's also not the behavior I
expect. The bug in the two .S files is very simple: identifying the
encoding method as "indirect pcrel" without actually putting in the
indirection. OK, what we need to do now is figure out what's different
between what you did and what I did.
I did my testing on a dual-2GHz G5 running OS X 10.3. I checked out
mainline (as of three days ago now), and did a clean 'make bootstrap'
with a new objdir, with "--enable-languages=c++,java". I then applied
my patch and did a new clean rebuild with a different objdir. I ran
'make -k check' at top level in the objdir with and without my patch,
and I compared the test results for libffi and libjava.
(Afterwards, when I saw that there were still failures in libjava, I
did another checkout, this time mainline as of March 8, and did yet
another clean bootstrap and 'make -k check' so I could compare.)
--Matt