Hi Eric,
On 2026-02-06 23:24, Eric Botcazou wrote:
I think I'm now seeing
+UNRESOLVED: gcc.dg/lto/pr122515 c_lto_pr122515_0.o-c_lto_pr122515.a
execute -f
lto=auto -ffat-lto-objects
+UNRESOLVED: gcc.dg/lto/pr122515 c_lto_pr122515_0.o-c_lto_pr122515.a link
-flto
=auto -ffat-lto-objects
+FAIL: gcc.dg/lto/pr122515, ar returned 1: /home/rguenther/obj/gcc/gcc-ar:
Cannot find plugin 'liblto_plugin.so'
because of this. This might be a pre-existing issue just exposed, but
it's still a bug.
Confirmed. What was the need for putting this on the 15 branch exactly?
I put it on the releases/gcc-15 branch to reduce number of failed tests.
Maybe I'm working for something that the community does not care about.
For me, I'd only want test cases failing if there are actual issues. In
this particular case, before my initial fix, the testing harness assumed
that all testing is performed in-tree. With my first patch, I fixed so that
the harness would allow tests to be performed both in- and out- of build
tree. Unfortunately, I lost the -B argument in the case of in-tree testing.
Regardless, my mistake affected a single test case, so I don't think it
was critical and I also corrected it within a day. I should, however, have
waited with pushing to releases/gcc-15 a day or two after my initial push
on trunk. Sorry for an problem this might have caused during the window of
my first and my 2nd fix.
Kind regards,
Torbjörn
