#6306 closed defect (fixed)
Random (unfrequent) failure on perl tests on Ubuntu Trusty
Reported by: | Even Rouault | Owned by: | Ari Jolma |
---|---|---|---|
Priority: | normal | Milestone: | |
Component: | PerlBindings | Version: | unspecified |
Severity: | normal | Keywords: | |
Cc: |
Description
Since I've added a Trusty target to our Travis-CI targets, I've noticed that the Perl tests fail from time to time (perhaps 5% of the builds ?), and specifically on that platform (Precise is fine)
Here's an example of such a failure : https://api.travis-ci.org/jobs/102071765/log.txt?deansi=true
PERL_DL_NONLAZY=1 /usr/bin/perl "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t t/00.t ......... ok t/01.t ......... ok t/02.t ......... ok t/03.t ......... All 10 subtests passed t/alg.t ........ ok t/callback.t ... ok t/error.t ...... ok t/gcp.t ........ ok t/gdal.t ....... ok t/nodata.t ..... ok t/ogr.t ........ ok t/open.t ....... ok t/osr.t ........ ok t/parenting.t .. ok t/utf8.t ....... ok t/util.t ....... ok Test Summary Report ------------------- t/03.t (Wstat: 139 Tests: 10 Failed: 0) Non-zero wait status: 139 Parse errors: No plan found in TAP output Files=16, Tests=1674, 7 wallclock secs ( 0.23 usr 0.02 sys + 2.16 cusr 0.20 csys = 2.61 CPU) Result: FAIL Failed 1/16 test programs. 0/1674 subtests failed. make[1]: *** [test_dynamic] Error 255 make[1]: Leaving directory `/home/travis/build/rouault/gdal_coverage/gdal/swig/perl' make: *** [test] Error 2
whereas correct runs (e.g https://api.travis-ci.org/jobs/102065636/log.txt?deansi=true) output :
PERL_DL_NONLAZY=1 /usr/bin/perl "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t t/00.t ......... ok t/01.t ......... ok t/02.t ......... ok t/03.t ......... ok t/alg.t ........ ok t/callback.t ... ok t/error.t ...... ok t/gcp.t ........ ok t/gdal.t ....... ok t/nodata.t ..... ok t/ogr.t ........ ok t/open.t ....... ok t/osr.t ........ ok t/parenting.t .. ok t/utf8.t ....... ok t/util.t ....... ok All tests successful. Files=16, Tests=1713, 2 wallclock secs ( 0.26 usr 0.01 sys + 2.36 cusr 0.28 csys = 2.91 CPU) Result: PASS
I've a Trusty system and ran
PERL_DL_NONLAZY=1 valgrind --trace-children=yes /usr/bin/perl "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t"
but it didn't emit any warning/error (other than memory leaks, but that's a different issue)
Change History (3)
comment:1 by , 9 years ago
comment:2 by , 9 years ago
Resolution: | → fixed |
---|---|
Status: | new → closed |
Fixed in r32965 for the test case but perhaps the possibility for a core dump in geotiff.cpp should also be fixed.
Well. The test 03.t is composed in a way which adds randomness to it (not very good). Basically the test, which causes the core dump is
"set the colorinterpretation of the band of a one band geotiff to 'RedBand'"
this leads to core dump in geotiff.cpp:4288 since there is no rasterband 2.