Skip Menu |

This queue is for tickets about the Cache-Profile CPAN distribution.

Report information
The Basics
Id: 129494
Status: open
Priority: 0/
Queue: Cache-Profile

People
Owner: Nobody in particular
Requestors: SREZIC [...] cpan.org
Cc:
AdminCc:

Bug Information
Severity: (no value)
Broken in: 0.06
Fixed in: (no value)



Subject: Failed test 'some cpu time accrued for set'
On some systems the test suite may fail: ... # Failed test 'some cpu time accrued for set' # at t/profile.t line 146. # '0' # >= # '0.001' # Looks like you failed 1 test of 91. t/profile.t ............ Dubious, test returned 1 (wstat 256, 0x100) Failed 1/91 subtests ... Sometimes it fails also for the "get" or "miss" test cases. On my smokers I see the failure more often on FreeBSD and CentOS systems, but never on Debian or Ubuntu.
On 2019-05-09 16:21:56, SREZIC wrote: Show quoted text
> On some systems the test suite may fail: > > ... > # Failed test 'some cpu time accrued for set' > # at t/profile.t line 146. > # '0' > # >= > # '0.001' > # Looks like you failed 1 test of 91. > t/profile.t ............ > Dubious, test returned 1 (wstat 256, 0x100) > Failed 1/91 subtests > ... > > Sometimes it fails also for the "get" or "miss" test cases. > > On my smokers I see the failure more often on FreeBSD and CentOS > systems, but never on Debian or Ubuntu.
Minor correction: it may fail on Debian wheezy (oldoldstable).
On 2019-05-09 13:21:56, SREZIC wrote: Show quoted text
> On some systems the test suite may fail:
I've seen this on and off in reports for a while now; I have no idea what the underlying cause is. If anyone has any insight, please let me know!
On 2019-05-09 20:41:47, ETHER wrote: Show quoted text
> On 2019-05-09 13:21:56, SREZIC wrote:
> > On some systems the test suite may fail:
> > I've seen this on and off in reports for a while now; I have no idea > what the underlying cause is. If anyone has any insight, please let me > know!
One possible problem: the accuracy for tracking CPU time might be quite low on some systems. I did some experiments: - CentOS 6 running in a VM: 10ms - FreeBSD 12 running in a VM: 8ms - Debian 10 running in a VM: ~100µs - Debian 8 running on hardware: ~50µs Given that the faster of the caching backends only need 7ms of real time for the set operation, it looks like it is too fast to accumulate CPU time on some systems. Maybe a good rule of thumb is: if the accumulated CPU time is exactly 0s, then the accuracy is too low and the test should probably be skipped.