Bug #124450 for Time-HiRes: 1.9754 breaks Catalyst-Runtime's test suite

Fri Feb 16 13:23:57 2018 SREZIC [...] cpan.org - Ticket created

CC:	jjnapiork [...] cpan.org
Subject:	1.9754 breaks Catalyst-Runtime's test suite

The test suite of Catalyst-Runtime-5.90117 seems to fail if the latest Time::HiRes is installed: # Failed test 'elapsed' # at t/unit_stats.t line 94. # got: '1518805116.2776' # expected: '14' # Looks like you failed 1 test of 13. t/unit_stats.t ........................................................ Dubious, test returned 1 (wstat 256, 0x100) Failed 1/13 subtests

Sat Feb 17 01:34:26 2018 JHI [...] cpan.org - Correspondence added

On Fri Feb 16 13:23:57 2018, SREZIC wrote: Show quoted text

> The test suite of Catalyst-Runtime-5.90117 seems to fail if the latest > Time::HiRes is installed: > > # Failed test 'elapsed' > # at t/unit_stats.t line 94. > # got: '1518805116.2776' > # expected: '14' > # Looks like you failed 1 test of 13. > t/unit_stats.t > ........................................................ > Dubious, test returned 1 (wstat 256, 0x100) > Failed 1/13 subtests

Uh, that result looks clearly like the the current epoch seconds. Is that Catalyst test using tv_interval()? That is about the only thing that changed for the 1.9754?

Sat Feb 17 01:34:27 2018 The RT System itself - Status changed from 'new' to 'open'

Sat Feb 17 02:54:40 2018 SREZIC [...] cpan.org - Correspondence added

RT-Send-CC:

jjnapiork [...] cpan.org, CAFEGRATZ [...] cpan.org

On 2018-02-17 01:34:26, JHI wrote: Show quoted text

> On Fri Feb 16 13:23:57 2018, SREZIC wrote:

> > The test suite of Catalyst-Runtime-5.90117 seems to fail if the > > latest > > Time::HiRes is installed: > > > > # Failed test 'elapsed' > > # at t/unit_stats.t line 94. > > # got: '1518805116.2776' > > # expected: '14' > > # Looks like you failed 1 test of 13. > > t/unit_stats.t > > ........................................................ > > Dubious, test returned 1 (wstat 256, 0x100) > > Failed 1/13 subtests

> > Uh, that result looks clearly like the the current epoch seconds. > > Is that Catalyst test using tv_interval()? That is about the only > thing that changed > for the 1.9754?

I'll check later. I found another "victim": CAFEGRATZ/Devel-TimeStats-0.04.tar.gz The test failure here looks similar: ... # Failed test 'elapsed' # at t/unit_stats.t line 100. # got: '1518820807.8718' # expected: '14' # Looks like you failed 1 test of 13. t/unit_stats.t .. Dubious, test returned 1 (wstat 256, 0x100) Failed 1/13 subtests ...

Sat Feb 17 07:32:43 2018 zefram [...] fysh.org - Correspondence added

Subject:	Re: [rt.cpan.org #124450] 1.9754 breaks Catalyst-Runtime's test suite
Date:	Sat, 17 Feb 2018 12:32:26 +0000
To:	Slaven_Rezic via RT <bug-Time-HiRes [...] rt.cpan.org>
From:	Zefram <zefram [...] fysh.org>

Both victims are using tv_interval(), in very similar manner, and crucially they're testing it in the same manner too. The test involves monkey-patching Time::HiRes::gettimeofday() to fake the time, and the modules make use of the one-arg form of tv_interval(). The old Perl tv_interval() when called with one arg would call Time::HiRes::gettimeofday() to get the current time, but the new XS tv_interval() calls the C gettimeofday() directly and so isn't affected by the monkey-patching. -zefram

Sat Feb 17 11:59:51 2018 JHI [...] cpan.org - Correspondence added

On Sat Feb 17 07:32:43 2018, zefram@fysh.org wrote: Show quoted text

> Both victims are using tv_interval(), in very similar manner, > and crucially they're testing it in the same manner too. The test > involves monkey-patching Time::HiRes::gettimeofday() to fake the > time, and the modules make use of the one-arg form of tv_interval(). > The old Perl tv_interval() when called with one arg would call > Time::HiRes::gettimeofday() to get the current time, but the new XS > tv_interval() calls the C gettimeofday() directly and so isn't affected > by the monkey-patching.

Thanks, Zefram! Then it doesn't sound like Time::HiRes can immediately do anything useful to cater for this expected behavior. Code expecting to monkey patch (or "virtualize") the time sources like this should probably use instead a code ref pointing to whatever. Show quoted text

> -zefram

Sun Feb 18 07:47:15 2018 RANDIR [...] cpan.org - Correspondence added

I've reported this issue to both affected modules.

Mon Feb 19 01:35:42 2018 SREZIC [...] cpan.org - Reference by ticket #124459 added

Mon Feb 19 01:35:43 2018 SREZIC [...] cpan.org - Reference by https://github.com/cafe01/Devel-TimeStats/issues/3 added