Another 28 XPASSes in jacks.xfail

Wed Oct 8 23:14:00 GMT 2003

Hi Tom,
> Roger> However on CVS mainline I'm still seeing 28 XPASSes on
> Roger> i686-pc-linux-gnu on a Red Hat 8.0 system. Are you seeing
> Roger> these?
>> I get 5 XPASSes in jacks right now:
>> 633:XPASS: 8.2-accessibility-inherited-member-5
> 1358:XPASS: 5.1.3-dti-1
> 1359:XPASS: 5.1.3-dti-2
> 1366:XPASS: 5.1.3-fti-1
> 1367:XPASS: 5.1.3-fti-2
>> This is a relatively-unmodified trunk gcj. (My one patch in the tree
> shouldn't affect jacks results at all.)
>> Roger> As I may have mentioned before I'm also seeing
> Roger> FAILs for tests that should be passing; 15.18.2-assoc-2 and
> Roger> 15.18.2-assoc-3.
>> These pass for me.
>> Do you have a patched gcj? If not, I wonder why our results would
> differ.

No, completely unpatched. The gcc tree I use for testing is a
clean check-out from mainline CVS. Very odd!
If its any help my "gcj_setup" script contains (after a "make -k check"):
set JAVAC /home/roger/GCC/JIKES/clean/gcc/gcj
set JAVA_CLASSPATH "/home/roger/GCC/JIKES/clean/i686-pc-linux-gnu/libjava/testsuite/../libgcj-3.4.jar"
set JAVAC_FLAGS {-B/home/roger/GCC/JIKES/clean/gcc/ --encoding=UTF-8 -C}
set JAVA /home/roger/GCC/JIKES/clean/i686-pc-linux-gnu//libjava/gij
set JAVA_FLAGS ""
set JAVAC_ENCODING_FLAG --encoding=
set tcltest::testConstraints(encoding) 1
set tcltest::testConstraints(gcj) 1
Atleast I now know not to try tracking down JACKS failures until I
can resolve this discrepancy. Is there a chance the tests are using
an incorrect (i.e. system) classpath or LD_LIBRARY_PATH? Is anyone
else seeing this?
Roger
--