New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
t/perf/benchmarks.t: premature exit on threaded builds on FreeBSD-13 #16752
Comments
From @jkeenanRecently we've been getting smoke-test failures on FreeBSD-13-CURRENT in http://perl5.test-smoke.org/report/74129 Note that the failures in t/perf/benchmarks.t are only occuring on I have not seen such failures on FreeBSD-11 (or any other platform, for So these failures could be due to some combination of (a) changes in To diagnose this I installed a FreeBSD-13 VM on a FreeBSD-11 host -- the I have been able to reproduce the failures when I build blead with these ##### $ cd t;./perl harness -v perf/benchmarks.t; cd - Test Summary Report perf/benchmarks.t (Wstat: 3072 Tests: 0 Failed: 0) However, when I configure without threads, all tests PASS: ##### That reproduces the results Carlos Guevara is reporting in the smoke When I conduct an internet search for "Cannot allocate memory while ##### In contrast, on the host on which this VM sits, the corresponding lines ##### This leads me to suspect that the problem is (c) -- a resource I experimented at splitting the fixtures in t/perf/benchmarks into two Unless someone has a better way of handling this problem, I'd like to Thank you very much. |
From @jkeenanSummary of my perl5 (revision 5 version 29 subversion 5) configuration: Characteristics of this binary (from libperl): |
From @iabynOn Sat, Nov 17, 2018 at 12:58:47PM -0800, James E Keenan (via RT) wrote:
At that point perf/benchmarks.t is simply executing do 'perf/benchmarks'; and perf/benchmarks is a 2000-ish line chunk of perl src which So either the VM doesn't have enough memory to compile that bit of code,
it's much more likely to be one of these: data seg size (kbytes, -d) unlimited What do those give you on the BSD VM and its host?
The file t/perf/benchmarks is the core of a benchmarking system for the We should probably just skip the test file when 'ulimit -X' gives a low -- |
The RT System itself - Status changed from 'new' to 'open' |
From @jkeenanOn Sat, 17 Nov 2018 22:37:12 GMT, davem wrote:
# On the host: ##### In the FreeBSD-13 VM: #####
-- |
From @eserteDana Sat, 17 Nov 2018 12:58:46 -0800, jkeenan@pobox.com reče:
Maybe related: ZEFRAM/XML-Easy-0.011.tar.gz fails with perl 5.29.4 with a warning WARNING: Complex regular subexpression recursion limit (65534) exceeded at t/syntax_main.t line 79, <GEN0> line 1635. (seen on linux + freebsd). This does not happen with 5.29.3. So maybe something in blead is consuming more resources? |
From @jkeenanOn Sun, 18 Nov 2018 16:52:44 GMT, slaven@rezic.de wrote:
No, this is a situation where a change in blead exposed sub-optimal code in XML::Easy's test suite. This was detected in the October "CPAN-river-3000" run and I provided a patch to the author. See: https://rt.cpan.org/Ticket/Display.html?id=127416#txn-1816119 Thank you very much. |
From @iabynOn Sat, Nov 17, 2018 at 02:44:15PM -0800, James E Keenan via RT wrote:
[ all other values identical on host and VM] That's odd. Does it fail if you execute the script directly: ./perl t/perf/benchmarks and if so does it appear to be just the size of the array of hashes it -- |
From @jkeenanOn Mon, 19 Nov 2018 14:55:04 GMT, davem wrote:
##### Test Summary Report perf/benchmarks.t (Wstat: 3072 Tests: 0 Failed: 0) Based on my debugging the other day, I would say that it's the size of the array of hashes rather than any specific element therein. I started by commenting out all tests, then uncommenting them one "block" at a time -- where a block is a group of elements with similar keys -- then running the test. Repeat until FAIL. I got about 70% of the way through the file before I got the error. I then reversed the process, commenting all tests out and uncommenting them from the bottom upwards. Once again, I got about 70% of the way through the file before the error occurred. -- |
From @iabynOn Mon, Nov 19, 2018 at 10:57:48AM -0800, James E Keenan via RT wrote:
I've just pushed the following commit. Can you retry ./perl -Ilib t/perf/benchmarks.t and show me what the error message is now? commit 516795a t/perf/benchmarks.t: improve do error checks -- |
From @jkeenanOn Tue, 20 Nov 2018 11:12:39 GMT, davem wrote:
This looks good. ##### ##### I will close this ticket once I see smoke-test results. Thank you very much. -- |
I should have checked the smoke results a year-and-a-half ago!
We're good. Closing. |
Migrated from rt.perl.org#133663 (status was 'open')
Searchable as RT133663$
The text was updated successfully, but these errors were encountered: