Bug #106304 for File-Path: Dependency not identified by multithreaded blead build

Sat Aug 08 08:50:51 2015 RICHE [...] cpan.org - Ticket created

Subject:

Dependency not identified by multithreaded blead build

See http://www.nntp.perl.org/group/perl.daily-build.reports/2015/08/msg180930.html File::Path is testing before Carp, which causes the build failure. ln 35425 of http://m-l.org/~perl/smoke/perl/win32/smoke-me/Karl%20Williamson/log55eb4084d561c28e746ac646230934efaced8327.log.gz shows the failure ln 36185 of the same file shows Carp being tested This may be a reflection of 'require Carp' in a sub

Sat Aug 08 09:21:21 2015 haarg [...] haarg.org - Correspondence added

perl itself builds differently than CPAN modules. All of the core modules are available. The problem here is that the paths in @INC are relative, but the test is changing directories before loading Carp.

Sat Aug 08 09:21:22 2015 The RT System itself - Status changed from 'new' to 'open'

Sat Aug 08 09:51:28 2015 RICHE [...] cpan.org - Correspondence added

On Sat Aug 08 09:21:21 2015, haarg wrote: Show quoted text

> perl itself builds differently than CPAN modules. All of the core > modules are available. > > The problem here is that the paths in @INC are relative, but the test > is changing directories before loading Carp.

Well, that is an awesome hint. I guess there's a few choices here. - Change how Carp is loaded (which I really don't want to touch) - Manipulate @INC to fq path in the test's BEGIN block (which I don't want to do because users could have relative @INC) - Change tests to not chdir ... but it changes a lot of test code - After each test block, change back to the original directory ... which probably is the best thing to do

Sat Aug 08 10:13:59 2015 jkeenan [...] cpan.org - Correspondence added

On Sat Aug 08 09:51:28 2015, RICHE wrote: Show quoted text

> On Sat Aug 08 09:21:21 2015, haarg wrote:

> > perl itself builds differently than CPAN modules. All of the core > > modules are available. > > > > The problem here is that the paths in @INC are relative, but the test > > is changing directories before loading Carp.

> > > Well, that is an awesome hint. I guess there's a few choices here. >

[snip] Show quoted text

> - After each test block, change back to the original directory ... > which probably is the best thing to do

Without having looked at Path.t, my hunch is that that is indeed the best thing to do. Rich: If you want to split up the workload on these test modifications, just let me know. Thank you very much. Jim Keenan

Sat Aug 08 11:10:51 2015 rich [...] richelberger.com - Correspondence added

Subject:	Re: [rt.cpan.org #106304] Dependency not identified by multithreaded blead build
Date:	Sat, 8 Aug 2015 11:10:36 -0400
To:	"bug-File-Path [...] rt.cpan.org" <bug-File-Path [...] rt.cpan.org>
From:	Richard Elberger <rich [...] richelberger.com>

I can start on it this evening and will see if it starts to become a chore. But I think it will be pretty quick once the pattern is identified. I do need to figure an effective way of testing this though. It would be ideal to repro it outside of a full on Perl build. Sent from my iPhone Show quoted text

> On Aug 8, 2015, at 10:13, James E Keenan via RT <bug-File-Path@rt.cpan.org> wrote: > > Queue: File-Path > Ticket <URL: https://rt.cpan.org/Ticket/Display.html?id=106304 > >

>>> On Sat Aug 08 09:51:28 2015, RICHE wrote: >>> On Sat Aug 08 09:21:21 2015, haarg wrote: >>> perl itself builds differently than CPAN modules. All of the core >>> modules are available. >>> >>> The problem here is that the paths in @INC are relative, but the test >>> is changing directories before loading Carp.

>> >> >> Well, that is an awesome hint. I guess there's a few choices here.

> [snip] >

>> - After each test block, change back to the original directory ... >> which probably is the best thing to do

> > Without having looked at Path.t, my hunch is that that is indeed the best thing to do. Rich: If you want to split up the workload on these test modifications, just let me know. > > Thank you very much. > Jim Keenan > >

Mon Aug 10 20:22:34 2015 jkeenan [...] cpan.org - Correspondence added

Examining the chart summarizing Perl 5 smoke testing at develop.help.org, the first Win2000 smoke run which reports a failure in cpan/File-Path/t/Path.t is this one: http://www.nntp.perl.org/group/perl.daily-build.reports/2015/07/msg180327.html ... which, not surprisingly, is the first report on this OS after Chris Williams merged File-Path 2.11 into blead. http://perl5.git.perl.org/perl.git/commit/139271cd3e9cb3cf25072ebdd400e52275c61b96

Wed Mar 08 16:25:05 2017 jkeenan [...] cpan.org - Correspondence added

On Sat Aug 08 08:50:51 2015, RICHE wrote: Show quoted text

> See > > http://www.nntp.perl.org/group/perl.daily- > build.reports/2015/08/msg180930.html > > File::Path is testing before Carp, which causes the build failure. > > ln 35425 of http://m-l.org/~perl/smoke/perl/win32/smoke- > me/Karl%20Williamson/log55eb4084d561c28e746ac646230934efaced8327.log.gz > shows the failure > > ln 36185 of the same file shows Carp being tested > > This may be a reflection of 'require Carp' in a sub

I *think* (but cannot prove) that this situation has cleared up. The problem was originally reported against CPAN version 2.11, but we have had 2.12 for a long time. When I go to matrix.cpantesters.org, I don't see any failures for File-Path 2.12_001 on MSWin32. See http://matrix.cpantesters.org/?dist=File-Path;os=mswin32;reports=1 which includes reports from testers who are usually quite reliable. I also don't see any failures for this test when it's run as part of smoke tests of Perl 5 blead -- but I have to qualify that by noting that at the present time we have only 1 steady source of smoke tests on Win32. See http://perl5.test-smoke.org/search and filter on 'MSWin32' for operating system. Rich, can you confirm that this has been resolved? Thank you very much. Jim Keenan

Sat Jun 03 17:08:22 2017 jkeenan [...] cpan.org - Correspondence added

On Wed Mar 08 16:25:05 2017, JKEENAN wrote: Show quoted text

> On Sat Aug 08 08:50:51 2015, RICHE wrote:

> > See > > > > http://www.nntp.perl.org/group/perl.daily- > > build.reports/2015/08/msg180930.html > > > > File::Path is testing before Carp, which causes the build failure. > > > > ln 35425 of http://m-l.org/~perl/smoke/perl/win32/smoke- > > me/Karl%20Williamson/log55eb4084d561c28e746ac646230934efaced8327.log.gz > > shows the failure > > > > ln 36185 of the same file shows Carp being tested > > > > This may be a reflection of 'require Carp' in a sub

> > I *think* (but cannot prove) that this situation has cleared up. The > problem was originally reported against CPAN version 2.11, but we have > had 2.12 for a long time. When I go to matrix.cpantesters.org, I > don't see any failures for File-Path 2.12_001 on MSWin32. See > http://matrix.cpantesters.org/?dist=File-Path;os=mswin32;reports=1 > which includes reports from testers who are usually quite reliable. > > I also don't see any failures for this test when it's run as part of > smoke tests of Perl 5 blead -- but I have to qualify that by noting > that at the present time we have only 1 steady source of smoke tests > on Win32. See http://perl5.test-smoke.org/search and filter on > 'MSWin32' for operating system. > > Rich, can you confirm that this has been resolved? > > Thank you very much. > Jim Keenan

File-Path 2.13's CPANtesters reports on Win32 are all green so far, so I am marking this ticket Resolved. Thank you very much. Jim Keenan

Sat Jun 03 17:08:38 2017 jkeenan [...] cpan.org - Status changed from 'open' to 'resolved'

Sat Jun 03 17:08:43 2017 jkeenan [...] cpan.org - Taken