Setting up Binutils workers/builders on Solaris

Rainer Orth ro@CeBiTec.Uni-Bielefeld.DE
Sat Jun 6 14:32:31 GMT 2026


Hi Frank,

>> * In my local testing, I've always run make -k -j<N> check as is
>>   customary to avoid failure of one testsuite to break the others.  This
>>   is what happens here: make check-gprof fails, so the two others
>>   (libctf, libsframe) aren't even run though they pass just fine
>>   locally.
>
> Agreed, -k makes sense, added.

I just saw: we're down to the gprof issue now.  Thanks.

>> * In my own testing, I've always used the toplevel mail-report.log
>>   target, which runs contrib/test_summary -t, to generate a easy to
>>   digest report of testsuite results, as is done over in gcc.  The gprof
>>   failures didn't show up there as FAILs, so this went unnoticed.
>
> FWIW, bunsen is a more thorough & useful sink for all the testsuite
> data (including non-dejagnu), and it's automatically collected from
> our buildbot runs.

Seems so: though I've been clamoring for the non-DejaGnu testsuites to
produce *.sum/*.log files for test_summary -t to pick up, it has only
been partially successful (libgo did it, e.g.).

I now see what's going on with the gprof tests: they all include main in
the counters, like

tst-gmon-gprof.sh ../gprof tst-gmon
--- expected
+++ actual
@@ -1,3 +1,4 @@
 f1 2000
 f2 1000
 f3 1
+main 1
FAIL

but am still trying to figure out what causes this.  However, that's a
matter for the binutils list.

	Rainer

-- 
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University


More information about the Buildbot mailing list