[Important] Major testing problems

It's a new year, and now we really need to get these problems with the regression testing system fixed. At least three test systems aren't getting their results posted on the results page. Two of these - Martin Wille and MetaComm - are extremely important to the health of Boost. The OSL3 and MetaComm testers seem to have uploaded incomplete .xml files. OSL3.xml, for instance, contains only the contents of comment.html and nothing else. The bjam log, however, gives more information. I've appended the bjam failure to the end of the message. My hunch is that the requirements for toolset::require-boost- spirit-support aren't getting propagated correctly. Perhaps the OSL3 tester should be configured for Spirit 1.6.x, but that shouldn't break testing completely, should it? As for MetaComm, I can only speculate what's happening there. I'm guessing it's either (1) the same problem that OSL3 is having, because MetaComm tests some compilers that can't handle the new spirit (e.g., Borland, MSVC <= 7); or (2) process_jam_log is failing to build because of the i18n changes in Filesystem. The Martin Wille tester looks like it is submitting complete XML, but its results pop in and out of existence. I don't even have a guess about what's happening here. Our testing has been broken for far to long. Let's fix it! Doug boost-test(COMPILE) "property_map/property_map_cc" : "libs/ property_map/test/property_map_cc.cpp" **** spirit 1.6x required to build library with this compiler **** **** skipping build of <@boost!libs!serialization! build>libboost_serialization.a; toolset= gcc-2.95.3-linux variant= debug **** /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:1585: in find-compatible-subvariant *** argument error * rule split-target-subvariant ( target-var properties-var toolset- var variant-var : subvariant ) * called with: ( sv-target sv-properties sv-toolset sv-variant : ) * missing argument subvariant /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:2032:see definition of rule 'split-target-subvariant' being called /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:1654: in link-libraries /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:1857: in generate-dependencies /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:1964: in subvariant-target /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:2057: in main-target /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:1369: in declare-local-target /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/testing.jam: 64: in boost-test /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/testing.jam: 552: in run ../libs/ptr_container/test/Jamfile:12: in sc-test ../libs/ptr_container/test/Jamfile:45: in load-jamfiles /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/ allyourbase.jam:1764: in SubInclude /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/ allyourbase.jam:1352: in subinclude Jamfile:48: in load-jamfiles /tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/ bootstrap.jam:21: in boost-build /tmp/dgregor/BoostRegressionTesting/boost/boost-build.jam:16: in module scope

Douglas Gregor wrote:
The Martin Wille tester looks like it is submitting complete XML, but its results pop in and out of existence. I don't even have a guess about what's happening here.
This has been fixed by Aleksey. See http://comments.gmane.org/gmane.comp.lib.boost.testing/2478 for details. Markus

Douglas Gregor wrote:
Our testing has been broken for far to long. Let's fix it!
/tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:1585: in find-compatible-subvariant *** argument error
Didn't even know this was a problem. Was there a previous mention of it I missed? ...Anyway it's fixed now, with a minor mod to libs/ptr_container/test/Jamfile. -- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim - Grafik/jabber.org

On Jan 2, 2006, at 11:58 AM, Rene Rivera wrote:
Douglas Gregor wrote:
Our testing has been broken for far to long. Let's fix it!
/tmp/dgregor/BoostRegressionTesting/boost/tools/build/v1/boost- base.jam:1585: in find-compatible-subvariant *** argument error
Didn't even know this was a problem. Was there a previous mention of it I missed? ...Anyway it's fixed now, with a minor mod to libs/ptr_container/test/Jamfile.
I mentioned it under the guise of Date-Time, because I thought it was specific to that library. Thanks for the fix! I'll restart the OSL3 tester to see if things have improved... Doug
participants (3)
-
Douglas Gregor
-
Markus Schöpflin
-
Rene Rivera