Bug #122699 for strictures: strictures enables fundamentaly broken fatal warnings and therefore is unusable

Thu Aug 03 08:54:26 2017 pali [...] cpan.org - Ticket created

Subject:

strictures enables fundamentaly broken fatal warnings and therefore is unusable

Same problem as for module common::sense reported in ticket: https://rt.cpan.org/Ticket/Display.html?id=122608 Due to broken fatal warnings in perl, and because strictures module enable broken fatal warnings, it is broken itself. Here is same example as from common::sense ticket: $ touch /tmp/test $ perl5.10.1 -MData::Dumper -e 'stat "/tmp/test"; eval { -l TEST; 1 }; print Dumper([stat _])' $VAR1 = []; $ touch /tmp/test $ perl5.10.1 -MData::Dumper -e 'use strictures 2; stat "/tmp/test"; eval { -l TEST; 1 }; print Dumper([stat _])' $VAR1 = [ 25, 2977687, 33188, 1, 1000, 1000, 0, 0, 1501764352, 1501764352, 1501764352, 4096, 0 ]; Please do not enable broken fatal warnings to make module usable. Thanks!

Thu Aug 03 15:20:18 2017 rt-cpan [...] trout.me.uk - Correspondence added

On Thu Aug 03 08:54:26 2017, PALI wrote: Show quoted text

> Same problem as for module common::sense reported in ticket: > https://rt.cpan.org/Ticket/Display.html?id=122608 > > Due to broken fatal warnings in perl, and because strictures module > enable broken fatal warnings, it is broken itself. Here is same > example as from common::sense ticket: > > $ touch /tmp/test > $ perl5.10.1 -MData::Dumper -e 'stat "/tmp/test"; eval { -l TEST; 1 }; > print Dumper([stat _])' > $VAR1 = []; > > $ touch /tmp/test > $ perl5.10.1 -MData::Dumper -e 'use strictures 2; stat "/tmp/test"; > eval { -l TEST; 1 }; print Dumper([stat _])' > $VAR1 = [ > 25, > 2977687, > 33188, > 1, > 1000, > 1000, > 0, > 0, > 1501764352, > 1501764352, > 1501764352, > 4096, > 0 > ]; > > Please do not enable broken fatal warnings to make module usable. > Thanks!

I fail to see the problem here - '-l TEST' is obviously bogus in this situation, so it die()ing immediately when the warning is emitted is exactly what strictures intends to enable. The fact that this has the side effect of stopping the broken code from clearing the stat() cache doesn't seem to me to be a problem. What, specifically, do you consider broken here?

Thu Aug 03 15:20:20 2017 The RT System itself - Status changed from 'new' to 'open'

Wed Aug 09 04:12:46 2017 pali [...] cpan.org - Correspondence added

On Thu Aug 03 15:20:18 2017, MSTROUT wrote: Show quoted text

> On Thu Aug 03 08:54:26 2017, PALI wrote:

> > Same problem as for module common::sense reported in ticket: > > https://rt.cpan.org/Ticket/Display.html?id=122608 > > > > Due to broken fatal warnings in perl, and because strictures module > > enable broken fatal warnings, it is broken itself. Here is same > > example as from common::sense ticket: > > > > $ touch /tmp/test > > $ perl5.10.1 -MData::Dumper -e 'stat "/tmp/test"; eval { -l TEST; 1 > > }; > > print Dumper([stat _])' > > $VAR1 = []; > > > > $ touch /tmp/test > > $ perl5.10.1 -MData::Dumper -e 'use strictures 2; stat "/tmp/test"; > > eval { -l TEST; 1 }; print Dumper([stat _])' > > $VAR1 = [ > > 25, > > 2977687, > > 33188, > > 1, > > 1000, > > 1000, > > 0, > > 0, > > 1501764352, > > 1501764352, > > 1501764352, > > 4096, > > 0 > > ]; > > > > Please do not enable broken fatal warnings to make module usable. > > Thanks!

> > I fail to see the problem here - '-l TEST' is obviously bogus in this > situation, so it die()ing immediately when the warning is emitted is > exactly what strictures intends to enable. > > The fact that this has the side effect of stopping the broken code > from clearing the stat() cache doesn't seem to me to be a problem. > > What, specifically, do you consider broken here?

As stated in 122608 this is just one example which shows that fatal warnings are broken. There are more problems. use strictures changes result of executed code which is broken behavior. Moreover if you run this code with use strictures under different versions of perl you got different result. Without use strictures I got same result with perl5.10 and perl5.24.

Wed Aug 09 23:04:44 2017 ether [...] cpan.org - Correspondence added

On 2017-08-09 01:12:46, PALI wrote: Show quoted text

> this is just one example which shows that fatal warnings are broken

Could you explain how it shows that? Why is not clearing the stat cache "broken"?

Thu Aug 10 05:36:49 2017 haarg [...] haarg.org - Correspondence added

Show quoted text

> As stated in 122608 this is just one example which shows that fatal > warnings are broken. There are more problems. use strictures changes > result of executed code which is broken behavior. Moreover if you run > this code with use strictures under different versions of perl you got > different result. Without use strictures I got same result with > perl5.10 and perl5.24.

The example you gave is extremely weak. Although I can follow your logic regarding stat, I don't think the change from the fatal warning actually makes the behavior incorrect. Having strictures change the behavior of the code by enabling fatal warnings is kind of the point, so you'll need a better justification.