diff options
Diffstat (limited to 'test/run_tests.pl')
| -rw-r--r-- | test/run_tests.pl | 23 |
1 files changed, 17 insertions, 6 deletions
diff --git a/test/run_tests.pl b/test/run_tests.pl index 0ed97b2ca9..4899356a6e 100644 --- a/test/run_tests.pl +++ b/test/run_tests.pl @@ -314,12 +314,23 @@ $ret = $harness->runtests(map { [ abs2rel($_, rel2abs(curdir())), basename($_) ] } sort { reorder($a) cmp reorder($b) } keys %tests); -# $ret->has_errors may be any number, not just 0 or 1. On VMS, numbers -# from 2 and on are used as is as VMS statuses, which has severity encoded -# in the lower 3 bits. 0 and 1, on the other hand, generate SUCCESS and -# FAILURE, so for currect reporting on all platforms, we make sure the only -# exit codes are 0 and 1. Double-bang is the trick to do so. -exit !!$ret->has_errors if (ref($ret) eq "TAP::Parser::Aggregator"); +# If this is a TAP::Parser::Aggregator, $ret->has_errors is the count of +# tests that failed. We don't bother with that exact number, just exit +# with an appropriate exit code when it isn't zero. +if (ref($ret) eq "TAP::Parser::Aggregator") { + exit 0 unless $ret->has_errors; + exit 1 unless $^O eq 'VMS'; + # On VMS, perl converts an exit 1 to SS$_ABORT (%SYSTEM-F-ABORT), which + # is a bit harsh. As per perl recommendations, we explicitly use the + # same VMS status code as typical C programs would for exit(1), except + # we set the error severity rather than success. + # Ref: https://perldoc.perl.org/perlport#exit + # https://perldoc.perl.org/perlvms#$? + exit 0x35a000 # C facility code + + 8 # 1 << 3 (to make space for the 3 severity bits) + + 2 # severity: E(rror) + + 0x10000000; # bit 28 set => the shell stays silent +} # If this isn't a TAP::Parser::Aggregator, it's the pre-TAP test harness, # which simply dies at the end if any test failed, so we don't need to bother |