develooper Front page | perl.perl5.porters | Postings from June 2020

Re: Windows Support [Was: Re: Announcing Perl 7]

Thread Previous | Thread Next
From:
Christian Walde
Date:
June 28, 2020 21:45
Subject:
Re: Windows Support [Was: Re: Announcing Perl 7]
Message ID:
op.0mxypd0jydyjqt@xenwide.cihq
On Sun, 28 Jun 2020 21:01:54 +0200, James E Keenan <jkeenan@pobox.com> wrote:

> On 6/28/20 1:23 PM, Christian Walde wrote:
>> On Fri, 26 Jun 2020 20:57:39 +0200, John Lightsey <john@nixnuts.net> wrote:
>>
>>> I'd love to see Perl 8 fix [...] Windows support.
>>
>> 2 things here:
>>
>> 1. Please don't phrase it in a way that implies Perl doesn't support
>> windows. I've been developing Perl on windows for over a decade and it
>> works.
>>
>
> Unfortunately, the data available to us suggests we have problems
> maintaining Perl on Windows in optimal condition.
>
> Go to http://perl5.test-smoke.org/search and plug in "MSWin32" to the
> OSname drop-down.  Observe a long list of FAILs.
>
> Now, I grant you that it only takes one unit test failure in one test
> file to grade the whole smoke-test run FAIL.  Nonetheless, we do a good
> job of getting PASSes on Linux and at least 3 of the BSDs.
>
> We need smoke-test reports preferably (a) from recent Windows releases
> for the business environment; and (b) from testers skilled enough in,
> and dedicated enough to, Windows to be able to diagnose failures and
> work with us to fix them.
>
> Thank you very much.
> Jim Keenan
>

Alright, i reenabled my smoker. Please look out for its first report in a few hours. :)

I think last time i had it running it was generating a bunch of PASSes and everything seemed fine and nobody ever contacted me about it even if stuff switched to red.

http://perl5.test-smoke.org/search?andnot_arch=0&selected_arch=&andnotsel_arch=0&report_arch=&selected_osnm=MSWin32&andnotsel_osnm=0&report_osnm=MSWin32&selected_host=DigitizedSqueak&andnotsel_host=0&report_host=DigitizedSqueak&andnot_osvs=0&selected_osvs=&andnotsel_osvs=0&report_osvs=&andnot_comp=0&selected_comp=&andnotsel_comp=0&config_comp=&selected_perl=all&andnot_cver=0&selected_cver=&andnotsel_cver=0&config_cver=&page=5

Here's the last PASS that machine sent: http://perl5.test-smoke.org/report/91097
Followed by the first FAIL after that: http://perl5.test-smoke.org/report/91170
And that machine's last fail before it went offline a while: http://perl5.test-smoke.org/report/97253

You're correct that in both of these only a single test was at fault, and particularly in the latter one, not even one that indicates core is broken.

In fact, it appears that Net::Ping's port test is just plain flaky on windows: http://matrix.cpantesters.org/?dist=Net-Ping%202.73 and in fact, the specific test causing the issue is also TODO'd on hpux|os390|freebsd, so it's probably more to do with the dist itself, and not anything with Windows.

I've created a pull request for that purpose here: https://github.com/rurban/Net-Ping/pull/20

Otherwise, how can i instruct my smoker to skip specific test files?

Also, you could just commit the proposed change there to blead for now.

That said, i still contend that the phrasing "fix windows support" is misapplied, as Perl the executable handily supports windows and doesn't need fixing but improvement.

-- 
With regards,
Christian Walde

Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About