Kenneth Belcourt wrote:
On Jul 30, 2014, at 5:14 PM, Adam Wulkiewicz
wrote: Barend Gehrels wrote:
For Boost.Spirit it seems to be the same. No results on master, but on certain platforms (those of j.c. bell), there are results (July 25). The same as Geometry. Interesting, in this case sublibs file was added 4 years ago. So either nobody noticed that tests are missing for 4 years or something else causes the problem. Assuming that it wasn't merged in master recently, or was it?
AFAIR there was some change in Config(?) made some time ago which also caused some problems with missing tests, e.g. in Geometry and Math. But this should be fixed now, it works in develop. May it be somehow related?
When I run: b2 --dump-tests status/ the tests are listed. And AFAIR the results are sent to the server with logs and everything.
So I'm guessing that it's something wrong on the server-side or it's a bug in the Regression, assuming that the scripts/program generating the HTML results from XMLs sent by the runners is in this module. Should I create a ticket or do you have some other ideas? The tests are in the tester composite xml file (e.g. Sandia-gcc-4.8.2.xml) so they’re getting built and trying to run. Why they don’t show up in html could be that test-type and test-program in the xml test-log are both blank.
So I went back and looked in the test results/bjam.log and found this:
testing.capture-output /scratch/boost_testing/results/boost/bin.v2/libs/geometry/test/core/access.test/gcc-4.8.2/debug/access.run
LD_LIBRARY_PATH="/sierra/sntools/SDK/compilers/gcc/4.8.2-RHEL6/bin:/sierra/sntools/SDK/compilers/gcc/4.8.2-RHEL6/lib:/sierra/sntools/SDK/compilers/gcc/4.8.2-RHEL6/lib32:/sierra/sntools/SDK/compilers/gcc/4.8.2-RHEL6/lib64:$LD_LIBRARY_PATH" export LD_LIBRARY_PATH
status=0 if test $status -ne 0 ; then echo Skipping test execution due to testing.execute=off exit 0 fi
Why is testing.execute is off?
Isn't it just some script which somewhere later should display/output this 'Skipping...' message and exit if status were not equal to 0 (probably meaning that the test wasn't built properly or something similar)? If the above is true I'm guessing that it should be in the log for all libraries.
Can you in master turn off all tests in geometry/test except for, say, core. And perhaps all tests in core/Jamfile.v2 except for the access test and let tests cycle to see if that one test will run? Maybe b2’s memory is getting thrashed?
I'm not sure if we can do this now. AFAIK the master branch is currently closed. Regards, Adam