Bug #115284 for Devel-NYTProf-Callgrind-TicksDiff: Random test falures (hash randomization?)

Sun Jun 12 03:32:38 2016 SREZIC [...] cpan.org - Ticket created

Subject:

Random test falures (hash randomization?)

See http://matrix.cpantesters.org/?dist=Devel-NYTProf-Callgrind-TicksDiff+0.04 for an overview of fail/pass reports --- starting with perl 5.18 there are random fail reports. This could be an indication that there's a problem caused by hash randomization.

Sun Jun 12 14:03:25 2016 SREZIC [...] cpan.org - Correspondence added

On 2016-06-12 03:32:38, SREZIC wrote: Show quoted text

> See http://matrix.cpantesters.org/?dist=Devel-NYTProf-Callgrind- > TicksDiff+0.04 for an overview of fail/pass reports --- starting with > perl 5.18 there are random fail reports. This could be an indication > that there's a problem caused by hash randomization.

Happens also with 0.05.

Sun Jun 12 15:11:46 2016 tech [...] tool.de - Correspondence added

Subject:	Re: [rt.cpan.org #115284] Random test falures (hash randomization?)
Date:	Sun, 12 Jun 2016 21:11:07 +0200
To:	bug-Devel-NYTProf-Callgrind-TicksDiff [...] rt.cpan.org
From:	Andreas Hernitscheck <tech [...] tool.de>

Its a strange thing. I use Moose and have an attribute "files" which sometimes seems to return nothing when using files() to access it. But only sometimes. Am 12.06.16 um 20:03 schrieb Slaven_Rezic via RT: Show quoted text

> Queue: Devel-NYTProf-Callgrind-TicksDiff > Ticket <URL: https://rt.cpan.org/Ticket/Display.html?id=115284 > > > On 2016-06-12 03:32:38, SREZIC wrote:

>> See http://matrix.cpantesters.org/?dist=Devel-NYTProf-Callgrind- >> TicksDiff+0.04 for an overview of fail/pass reports --- starting with >> perl 5.18 there are random fail reports. This could be an indication >> that there's a problem caused by hash randomization.

> Happens also with 0.05. >

Sun Jun 12 15:11:46 2016 The RT System itself - Status changed from 'new' to 'open'

Thu Jun 16 16:09:26 2016 ANDK [...] cpan.org - Correspondence added

For what it is worth: I just let my smoker test with both v5.17.5-518-g7dc8663 and v5.17.5-517-g867b16b. The "517" sent a pass, the "518" (which is the one that introduced hash randomization) a fail. It's pretty obviously a hash randomization influence that triggers something here.

Fri Jun 17 04:02:22 2016 tech [...] tool.de - Correspondence added

Subject:	Re: [rt.cpan.org #115284] Random test falures (hash randomization?)
Date:	Fri, 17 Jun 2016 08:02:07 +0000
To:	bug-devel-nytprof-callgrind-ticksdiff [...] rt.cpan.org
From:	"Andreas Hernitscheck" <tech [...] tool.de>

I understand. For me it looks like the issue is part of Moose then. I have to investigate. On 16 June, 2016 22:09 Berlin, "Andreas Koenig via RT" <bug-Devel-NYTProf-Callgrind-TicksDiff@rt.cpan.org> wrote: Queue: Devel-NYTProf-Callgrind-TicksDiff Ticket <URL: [1]https://rt.cpan.org/Ticket/Display.html?id=115284 > For what it is worth: I just let my smoker test with both v5.17.5-518-g7dc8663 and v5.17.5-517-g867b16b. The "517" sent a pass, the "518" (which is the one that introduced hash randomization) a fail. It's pretty obviously a hash randomization influence that triggers something here. -- Andreas Hernitscheck References Visible links 1. https://rt.cpan.org/Ticket/Display.html?id=115284

Sat Jun 18 07:34:32 2016 SREZIC [...] cpan.org - Correspondence added

RT-Send-CC:

tech [...] tool.de

Possible reason: random order when initializing parameters. At least https://metacpan.org/pod/distribution/Moose/lib/Moose/Manual/Attributes.pod#Laziness mentions that order is not predictable in some situations. On 2016-06-17 04:02:22, tech@tool.de wrote: Show quoted text

> I understand. For me it looks like the issue is part of Moose then. I > have to investigate. > > > On 16 June, 2016 22:09 Berlin, "Andreas Koenig via RT" > <bug-Devel-NYTProf-Callgrind-TicksDiff@rt.cpan.org> wrote: > > Queue: Devel-NYTProf-Callgrind-TicksDiff > Ticket <URL: [1]https://rt.cpan.org/Ticket/Display.html?id=115284 > > > For what it is worth: I just let my smoker test with both > v5.17.5-518-g7dc8663 and v5.17.5-517-g867b16b. The "517" sent a pass, > the "518" (which is the one that introduced hash randomization) a fail. > It's pretty obviously a hash randomization influence that triggers > something here. > > > -- > Andreas Hernitscheck > > tech@tool.de > Helene-Weber-Allee 10 > 80637 Munich, Germany > > Phone: +49 176 210 16 397 > > References > > Visible links > 1. https://rt.cpan.org/Ticket/Display.html?id=115284 >

Sat Jun 18 09:50:22 2016 ANDK [...] cpan.org - Correspondence added

So I edited TicksDiff.pm and made all attributes lazy and with that all tests seem to pass. I tested only with one perl (5.20.3) and only for a few minutes in an endless loop, but I think this should be a valid solution.

Thu Jul 28 17:34:53 2016 SREZIC [...] cpan.org - Broken in 0.05 added