perl 5.42.0 build: Possibly Async 0.14 test will hang

[twata1]

Hello,

I apologise if this is just a problem with my environment.

It seems to me that Async version 0.14 test hangs.

Sample output:

C:\home\mintdev001>cpan Async

(snip)

Running make for A/AR/ARISTOTLE/Async-0.14.tar.gz
CPAN: Module::CoreList loaded ok (v5.20250702)
cp Async.pm blib\lib\Async.pm
(C:\strawberry-perl-5.42.0.1-beta1-64bit-portable\c\bin\gmake.exe exited with 0)
CPAN::Reporter: gmake result is 'pass', No errors.
  ARISTOTLE/Async-0.14.tar.gz
  C:\strawberry-perl-5.42.0.1-beta1-64bit-portable\c\bin\gmake.exe -- OK
Running make test for ARISTOTLE/Async-0.14.tar.gz
"C:\strawberry-perl-5.42.0.1-beta1-64bit-portable\perl\bin\perl.exe" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test
::Harness::Switches; test_harness(0, 'blib\lib', 'blib\arch')" t/*.t
# Testing on Perl 5.042000 at C:\strawberry-perl-5.42.0.1-beta1-64bit-portable\perl\bin\perl.exe
# Testing Async @ 0.14
t/00-load.t .............. ok
(hangs)

Perl version:

C:\home\mintdev001>perl -v

This is perl 5, version 42, subversion 0 (v5.42.0) built for MSWin32-x64-multi-thread

Copyright 1987-2025, Larry Wall

Perl may be copied only under the terms of either the Artistic License or the
GNU General Public License, which may be found in the Perl 5 source kit.

Complete documentation for Perl, including FAQ lists, should be found on
this system using "man perl" or "perldoc perl".  If you have access to the
Internet, point your browser at https://www.perl.org/, the Perl Home Page.


C:\home\mintdev001>

Thank you,

[sisyphus]

The tests hang for me at various points with all of my own recent builds of perl.
The PASSES from Windows on the testers matrix seem to be 32-bit builds. (I didn't check all of them.)
It's not a Strawberry-specific problem - more a general Win32 issue. (Even my 32-bit build of perl-5.42.0 hangs during the tests.)

My MSVC143 build of 32-bit 5.42.0 builds Async-0.14 fine ... so there's some hope.
But my MSVC143 build of 64-bit 5.42.0 suffers a hang in t/read-multiple.t.

It's a pure-perl module, and I have the feeling it might not be all that difficult for a sufficiently knowledgeable person to address.
(But then everything is solvable for a "sufficiently knowledgeable" person ;-)

At https://github.com/ap/Async there's the option (a tab) to submit a Pull Request, but no tab to create an Issue.
That suggests to me that the author, while prepared to apply fixes, is not interested in devising a solution to any problems that might arise.
Maybe try contacting the author privately. He's still quite active in the perl community - a member of the Perl Steering Committee, UIM.

[shawnlaffan]

Thanks Rob. That also means I don't need to worry about this for the 5.42 build which I am hoping to upload tomorrow.

At https://github.com/ap/Async there's the option (a tab) to submit a Pull Request, but no tab to create an Issue.

Metacpan lists the issue tracker as being on RT.

Some possibly relevant info about the history of the module is in https://rt.cpan.org/Public/Bug/Display.html?id=139416

[twata1]

@sisyphus @shawnlaffan

Thank you for your response.

Regarding my MSVC143 build of 64-bit, this issue appears to have started with Perl 5.41.2.

Sometimes, after repeated testing, it would pass.
Feel free to close this ticket.

Thank you,

[sisyphus]

Sometimes,t/destroy-siblings.t after repeated testing, it would pass.

I had a bit of a play around with it again and noticed the same thing.
I downloaded the source and built it manually on my MSWin32-x64-multi-thread gcc-build of perl-5.40.0. I repeatedly ran the "make test" step, which ran successfully half the time. For the other "half" it hanged at or near the end of t/destroy-siblings.t.
Maybe some race condition ?

As Shawn pointed out, the place for a bug report is the rt.cpan bug-tracker (https://rt.cpan.org/Public/Dist/Display.html?Name=Async).
I'm rather amazed that anyone could prefer the clunky and annoying old rt.cpan tracker over github.

[twata1]

Regarding my MSVC143 build of 64-bit, this issue appears to have started with Perl 5.41.2.

Hmm. It seems to happen in this version too.

@sisyphus

Maybe some race condition ?

I probably think so too.

Thank you,

[sisyphus]

Hmm. It seems to happen in this version too.

I'm wondering if it's related to UCRT.
Can you find a case where the hanging occurs when perl has been built to use MSVCRT ?
Can you find a case where it never occurs when perl has been built to use UCRT ?
On mingw-w64 builds of perl, if perl -V:libc reports '-lucrt' then it's using the UCRT. Otherwise it's using MSVCRT.
For MSVC builds, I know that MSVC143 uses UCRT, but I don't know about MSVC142 or earlier.

My cygwin builds of perl seem to be unaffected.
Another thing I've noticed is that the test suite always runs smoothly if, at the end of destroy-globals.t, destroy-siblings.t, error-read-blocking.t and read-multiple.t, I add the code:

sleep(1) if $^O =~ /MSWin/;

Is that sufficient for the test suite to always succeed for you ?

I'll do a bug report about this (and provide a link to that report here).
I'm thinking it might be a bug in the UCRT builds of Windows, in which case the issue should be reported to https://github.com/Perl/perl5/issues.

But I'll have a bit more of a fiddle about with it, and probably start with a ticket at Async's rt.cpan tracker. (See what Aristotle thinks.)

[twata1]

Can you find a case where the hanging occurs when perl has been built to use MSVCRT ?

Yes. Please see below. (If it hangs, press CTRL+C.)

C:\home\vega5420rc3\Async-0.14>gmake test
"C:\WinLibs\Perl-5.42.0-rc3\bin\perl.exe" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_h
arness(0, 'blib\lib', 'blib\arch')" t/*.t
t/00-load.t .............. # Testing on Perl 5.042000 at C:\WinLibs\Perl-5.42.0-rc3\bin\perl.exe
# Testing Async @ 0.14
t/00-load.t .............. ok
t/destroy-globals.t ...... ok
t/destroy-siblings.t ..... ok
t/error-fork.t ........... ok
t/error-pipe.t ........... ok
t/error-read-blocking.t .. ok
t/error-read-polling.t ... ok
t/read-multiple.t ........ ok
All tests successful.
Files=8, Tests=38,  3 wallclock secs ( 0.09 usr +  0.08 sys =  0.17 CPU)
Result: PASS

C:\home\vega5420rc3\Async-0.14>gmake test
"C:\WinLibs\Perl-5.42.0-rc3\bin\perl.exe" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_h
arness(0, 'blib\lib', 'blib\arch')" t/*.t
t/00-load.t .............. # Testing on Perl 5.042000 at C:\WinLibs\Perl-5.42.0-rc3\bin\perl.exe
# Testing Async @ 0.14
t/00-load.t .............. ok
t/destroy-globals.t ...... ok
t/destroy-siblings.t ..... ok
t/error-fork.t ........... ok
t/error-pipe.t ........... ok
t/error-read-blocking.t .. ok
t/error-read-polling.t ... ok
t/read-multiple.t ........ ok
All tests successful.
Files=8, Tests=38,  2 wallclock secs ( 0.08 usr +  0.08 sys =  0.16 CPU)
Result: PASS

C:\home\vega5420rc3\Async-0.14>gmake test
"C:\WinLibs\Perl-5.42.0-rc3\bin\perl.exe" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test::Harness::Switches; test_h
arness(0, 'blib\lib', 'blib\arch')" t/*.t
t/00-load.t .............. # Testing on Perl 5.042000 at C:\WinLibs\Perl-5.42.0-rc3\bin\perl.exe
# Testing Async @ 0.14
t/00-load.t .............. ok
t/destroy-globals.t ...... ok
t/destroy-siblings.t ..... 1/4 Terminating on signal SIGINT(2)
Terminate batch job (Y/N)? y
gmake: *** [makefile:851: test_dynamic] Error 255

C:\home\vega5420rc3\Async-0.14>perl -V:libc
libc='-lmsvcrt';

C:\home\vega5420rc3\Async-0.14>

Can you find a case where it never occurs when perl has been built to use UCRT ?

No.

C:\home\mintdev001\Async-0.14>gmake test
"C:\strawberry-perl-5.42.0.1-beta1-64bit-portable\perl\bin\perl.exe" "-MExtUtils::Command::MM" "-MTest::Harness" "-e" "undef *Test
::Harness::Switches; test_harness(0, 'blib\lib', 'blib\arch')" t/*.t
t/00-load.t .............. # Testing on Perl 5.042000 at C:\strawberry-perl-5.42.0.1-beta1-64bit-portable\perl\bin\perl.exe
# Testing Async @ 0.14
t/00-load.t .............. ok
t/destroy-globals.t ...... 1/15 Terminating on signal SIGINT(2)
Terminate batch job (Y/N)? y
gmake: *** [makefile:850: test_dynamic] Error 255

C:\home\mintdev001\Async-0.14>perl -V:libc
libc='-lucrt';

C:\home\mintdev001\Async-0.14>

Is that sufficient for the test suite to always succeed for you ?

MSVCRT (perl 5.42.0 rc3 built with gcc 14.2.0): Yes. I was surprised that inserting sleep into each test file caused the tests to pass.
UCRT (perl 5.42.0 rc2 built with MSVC143): Yes.
UCRT (strawberry perl 5.42.0.1 beta1 64bit portable): No.

P.S.
Since I went to the trouble of making it, I'll attach my ugly patch.
(Edit: This patch seems to reduce hangs, but it still appears to be incomplete.)
Async-0.14-for-win.patch

Thank you,

[sisyphus]

Ok - thanks @twata1.
Bug report has been lodged at https://rt.cpan.org/Ticket/Display.html?id=168826&results=a680b1faa55bd32555255119a28dae54

[twata1]

@sisyphus
You're welcome.
Personally, based on the symptoms alone, I think it resembles the following. (It may not be the same cause, though.)

noxxi/p5-io-socket-ssl#30

Thank you,