jkeenan
commented
3 years ago Thank you for this report. The operating system -- s390x -- for which you are attempting to build an RPM is one that we do not have access to and from which we have not received smoke-testing reports. Hence, we'll need some direct interaction with you in order to resolve the problem. I'd like to begin by posing these questions:
-
Can you reproduce the error if you simply repeat your process?
-
You appear to be patching an existing
Makefile.PLrather than simply using theMakefile.PLprovided in the CPAN distribution. Can you say why you are doing so? Can you provide the patch? (I ask because we have to rule out the possibility that this failure was introduced by something over which I have no control, e.g., your RPM build process.) -
As best I can tell from your build.log, the failure reported is this:
t/00-load.t .............. ok Unable to open '/home/jkeenan/gitwork/zzzothers/devel-nytprof/t/test01.p' for reading: No such file or directory t/11-reader.t ............ ok t/12-data.t .............. Dubious, test returned 1 (wstat 256, 0x100) Failed 1/79 subtests # Failed test 'Nothing dumped, as requested' # at t/12-data.t line 272.a. My hunch is that the
Unable to open ... t/test01.p for readingwarning is transient and not associated with the test failure. However, that's just a hunch based on observation of this warning in smoke-testing reports.
b. The failure you are getting comes from this part of t/12-data.t:
262 {
263 SKIP: {
264 skip "Bad interaction when trace_level is set", 3
265 if trace_level();
266 my $profile;
267 local $ENV{NYTPROF_ONLOAD} = 'alpha=beta:gamma=delta:dump=0';
268 my $stderr = capture_stderr {
269 $profile = Devel::NYTProf::Data->new( { filename => $file, quiet => 1 } );
270 };
271 ok(! $stderr, "Nothing dumped, as requested");
272 ok(defined $profile, "Direct call of constructor returned defined value");
273 isa_ok($profile, 'Devel::NYTProf::Data');
274 }
275
276 }What the test at line 271 is asserting is that, (a) when creating a profile of a Perl process (technically, creating a new Devel::NYTProf::Data object), (b) if we have set an environmental variable $NYTPROF_ONLOAD to not request dumping and (c) we capture the STDERR from a call to Devel::NYTProf::Data->new(), we should expect to capture no STDERR (at most, an empty string).
It would therefore be helpful to see how, in your test run, $stderr is being populated. Would it be possible for you to insert the following debugging statement after line 271 of t/12-data.t, re-run your process, and provide the output?
271 ok(! $stderr, "Nothing dumped, as requested");
print STDERR "<$stderr>\n" if $stderr;
272 ok(defined $profile, "Direct call of constructor returned defined value");My hunch is that, from the point of view of users of Devel-NYTProf from your RPM, the failure you are reporting is a false negative. The relevant part of lib/Devel/NYTProf/Data.pm is this:
135 # a hack for testing/debugging
136 # $ENV{NYTPROF_ONLOAD} must be a colon-delimited string of
137 # equal-sign-delimited substrings, e.g.,
138 # 'alpha=beta:gamma=delta:dump=1:exit=1';
139
140 if (my $env = $ENV{NYTPROF_ONLOAD}) {
141 my %onload = map { split /=/, $_, 2 } split /:/, $env, -1;
142 warn _dumper($profile) if $onload{dump};
143 exit $onload{exit} if defined $onload{exit};
144 }Hence, what line 271 of t/12-data.t is exercising is a part of the source code that is only of interest to NYTProf developers -- not to NYTProf end-users. If this proves to be the case, then (a) you could simply do the equivalent of --force install to override the test result; and/or (b) we could, in an upcoming release of the distribution, make this test PERL_AUTHOR_TESTING only.
We look forward to your response.
Thank you very much. Jim Keenan
jplesnik
I want to build the version 6.08 for Fedora, but the test 12-data.t fails on s390x
Complete build informati can be found on: https://koji.fedoraproject.org/koji/taskinfo?taskID=66958623