Subject: | smokecurrent.skiptests causes failures in t/porting/manifest.t |
Date: | Mon, 19 Nov 2018 20:56:11 -0500 |
To: | bug-Test-Smoke [...] rt.cpan.org |
From: | James E Keenan <jkeenan [...] pobox.com> |
Entering a test file in smokecurrent.skiptests appears to cause failures
in t/porting/manifest.t and perhaps other files.
Today I smoke-tested OpenBSD-6.4. My instance of that OS is a stock
virtualbox VM provided at vagrantup.com. Like many such boxes, when I
run the Perl 5 test suite or smoke test, I get a higher than normal
incidence of test failures attributable to resource constraints. So
today I tried something new: I made an entry to skip one test file in
smokecurrent.skiptests (and a corresponding entry in smokecurrent.usernote):
#####
[p5smoke] 45 $ cat smokecurrent.skiptests
# One test name on a line
cpan/Sys-Syslog/t/syslog.t
[p5smoke] 46 $ cat smokecurrent.usernote
cpan/Sys-Syslog/t/syslog.t has too many intermittent
false negatives. Skipping on this platform. See
https://rt.cpan.org/Ticket/Display.html?id=127454.
#####
The smoke test report generated was;
#####
http://perl5.test-smoke.org/report/74401
#####
cpan/Sys-Syslog/t/syslog.t was apparently skipped -- but
t/porting/manifest.t FAILed in every one of the 8 cells of the smoke test.
#####
v5.29.4-93-g8f6aacbbe9
---------------------- ----------------------------------------------------
F F F F (*1)
F F F F -Duseithreads (*1)
| | | |
| | | +------------ perlio DEBUGGING
| | +--------------- stdio DEBUGGING
| +------------------ perlio
+--------------------- stdio
Applied patches:
SMOKE8f6aacbbe92285ffc82603b952f356914d751df0
Skipped tests:
# One test name on a line
cpan/Sys-Syslog/t/syslog.t
Test failures:
~~ ../t/porting/manifest.t ..................................... FAILED
10364
[stdio/perlio]
[stdio/perlio] -Duseithreads DEBUGGING
[stdio/perlio] DEBUGGING
[stdio/perlio] -Duseithreads
~~ ../cpan/ExtUtils-Install/t/Installed.t ...................... FAILED
Non-zero exit status: 2 Bad plan. You planned 73 tests but ran 24.
[stdio] -Duseithreads DEBUGGING
...
cpan/Sys-Syslog/t/syslog.t has too many intermittent
false negatives. Skipping on this platform. See
https://rt.cpan.org/Ticket/Display.html?id=127454.
#####
Note also that cpan/ExtUtils-Install/t/Installed.t terminated
prematurely as well.
I went to the perl-current/ directory and manually ran the two test
files through the harness several times. They consistently PASSed.
I then examined the corresponding log file; see excerpt attached. The
failures in t/porting/manifest.t are clearly associated with the
presence of an entry in smokecurrent.skiptests:
#####
# Failed test 10364 - git ls-files gives the same number of files as
MANIFEST lists at porting/manifest.t line 95
# got "6117"
# expected "6116"
# Failed test 10366 - Nothing added to the repo that isn't in MANIFEST
at porting/manifest.t line 107
# got "1"
# expected "0"
# Failed test 10367 - Nothing added to the repo that isn't in MANIFEST
at porting/manifest.t line 108
# got "not in MANIFEST: cpan/Sys-Syslog/t/syslog.t"
# expected "not in MANIFEST: "
# parser guessed wrong encoding expected 'CP1252' got 'UTF-8'
# Will use Digest::MD5
*** Error 3 in /home/jkeenan/p5smoke/perl-current (makefile:829
'test_harness')
error while running harness target 'test_harness': 1 at
/home/jkeenan/p5smoke/lib/Test/Smoke/App/RunSmoke.pm line 118.
[2018-11-19 16:47:37-0500] [known failed test]
../t/porting/manifest.t.....................................
[2018-11-19 16:47:38-0500]
[2018-11-19 16:47:37-0500] [known failed test]
../t/porting/manifest.t.....................................
[2018-11-19 16:47:38-0500]
[2018-11-19 16:47:38-0500]
../t/porting/manifest.t.....................................FAILED
[2018-11-19 16:47:38-0500] 10364, 10366-10367
#####
Note that while the first failing test number listed is 10364, in a
normal run of this file that would be 10365.
#####
ok 10363 - Line 6117 doesn't start with a blank
ok 10364 - File zaphod32_hash.h exists
ok 10365 - MANIFEST sorted properly
ok 10366 - git ls-files gives the same number of files as MANIFEST lists
ok 10367 - no duplicate files in MANIFEST
ok 10368 - Nothing added to the repo that isn't in MANIFEST
ok 10369 - Nothing added to the repo that isn't in MANIFEST
ok 10370 - Nothing in the MANIFEST that isn't tracked by git
ok 10371 - Nothing in the MANIFEST that isn't tracked by git
#####
I then went to the Test-Smoke website matrix for failing tests and
examined
http://perl5.test-smoke.org/submatrix?test=../t/porting/manifest.t. I
spot-checked the reports for t/porting/manifest.t. Every one that I
looked at had an entry in smokecurrent.skiptests. See, e.g.,
http://perl5.test-smoke.org/report/70865 (from tester abeltje!):
#####
Skipped tests:
t/porting/pending-author.t
Test failures:
~~ ../t/porting/test_bootstrap.t ............................... FAILED
Non-zero exit status: 2
[stdio/perlio/locale:nl_NL.utf8] -Dcc='ccache gcc'
[stdio/perlio/locale:nl_NL.utf8] -Dcc='ccache gcc' -Duseithreads
DEBUGGING
[stdio/perlio/locale:nl_NL.utf8] -Dcc='ccache gcc' DEBUGGING
[stdio/perlio/locale:nl_NL.utf8] -Dcc='ccache gcc' -Duseithreads
~~ ../t/porting/manifest.t ..................................... FAILED 9824
[stdio/perlio/locale:nl_NL.utf8] -Dcc='ccache gcc'
[stdio/perlio/locale:nl_NL.utf8] -Dcc='ccache gcc' -Duseithreads
DEBUGGING
[stdio/perlio/locale:nl_NL.utf8] -Dcc='ccache gcc' DEBUGGING
[stdio/perlio/locale:nl_NL.utf8] -Dcc='ccache gcc' -Duseithreads
#####
(IMO, there is no need to skip t/porting/pending-author.t. In a normal
run of 'make test_harness', it is skipped anyway with message,
"porting/pending-author.t .. skipped: No pending changes".)
So as it currently stands, using 'smokecurrent.skiptests', rather than
avoiding false negatives, creates new ones.
Can you investigate?
Thank you very much.
Jim Keenan
Message body is not shown because sender requested not to inline it.