PERL Modules - ActiveState module failures

Author

ActiveState module failures

plugh

2005-03-15, 3:58 am

Hello Group,

Browsing through http://ppm.activestate.com/BuildStatus/, I see an
awful lot of modules (both Tk and non) classified as FAILs. More
FAILs than PASSes, looks to me.

Looking at some of the failure messages, I see a high percentage
of things like "Can't locate Test/More.pm" and "couldn't connect
to display :0".

Seems to me these should be called tester failures rather than
module failures, no? Does anyone know what's going on there?

Regards,

plugh

Sisyphus

2005-03-15, 3:58 am

plugh wrote:
> Hello Group,
>
> Browsing through http://ppm.activestate.com/BuildStatus/, I see an
> awful lot of modules (both Tk and non) classified as FAILs. More
> FAILs than PASSes, looks to me.
>
> Looking at some of the failure messages, I see a high percentage
> of things like "Can't locate Test/More.pm" and "couldn't connect
> to display :0".
>
> Seems to me these should be called tester failures rather than
> module failures, no? Does anyone know what's going on there?
>

AS have an automated build procedure. If the module builds without
error, then they build a ppm of it - else it fails to build and they
simply report that it failed. It's then up to the author of the module
to attend to the problem.

I have no idea why Test::More would be unlocatable on their build system.

Cheers,
Rob

--
To reply by email u have to take out the u in kalinaubears.

Sisyphus

2005-03-16, 8:56 pm

plugh wrote:

>
> If it were only a few bogus FAILs, I wouldn't be concerned, but a
> casual look at a selection of failures indicates that perhaps 50%
> of all the failed modules (maybe 25% of ALL the modules) are marked
> FAIL when they may actually be OK.
>

Can you give some specific examples of modules that look as though they
should have passed ? - ones that failed because Test::More was
unlocatable would be a good starting point.

I'm sufficiently curious to take a look at this, and let them know if
there is indeed a problem with their build system. But I couldn't find
any examples of the type you mentioned. (I've got a pretty slow
connection here, and didn't try a large sample.)

Also, what OS are we concerned with here ? I assume it's Win32.

Cheers,
Rob

--
To reply by email u have to take out the u in kalinaubears.

plugh

2005-03-18, 3:56 am

On Wed, 16 Mar 2005 22:46:38 +0000, Sisyphus wrote:

> plugh wrote:
>
>
> Can you give some specific examples of modules that look as though they
> should have passed ? - ones that failed because Test::More was unlocatable
> would be a good starting point.
>
> [ . . . ]
> Also, what OS are we concerned with here ? I assume it's Win32.

Well, I was looking more at the linux and solaris columns - I don't
use DOS. I know AS is mostly Windows oriented, but they're including test
results for other OSes too.

Here's a selection of suspicious reports from the 5.8 A section. I didn't
run across a missing Test::More this time (hmm, maybe I saw those in the
5.6 section?), but there are plenty of other things. I'll concentrated on
modules that fail across the board (for all or most of the OSes):

Here are two that look like everything passed, but they're listed as a
FAILs:
http://ppm.activestate.com/BuildSta...Correct-0.4.txt
http://ppm.activestate.com/BuildSta...orizer-0.07.txt

Here's a 'can't chown test module: Operation not permitted' - could be a
module problem, but I'd suspect the test setup first:
http://ppm.activestate.com/BuildSta...-Abuse-0.02.txt

These modules "Can't locate Module/Build.pm":
http://ppm.activestate.com/BuildSta...senham-0.10.txt
http://ppm.activestate.com/BuildSta...nBucket-0.2.txt
Looks like a bunch more get stuck in some configuration loop trying to
build Module/Build until they eventually bail and fail. Here's one of
those (caution: it's a *long* listing):
http://ppm.activestate.com/BuildSta...p-Info-0.44.txt

I dunno what the heck happened on this one, but again I suspect the test
setup first:
http://ppm.activestate.com/BuildSta...rksort-1.05.txt

I also see a lot of "Can't open perl script "Makefile.PL": No such file or
directory". For example:
http://ppm.activestate.com/BuildSta...y-PAT-1.0.0.txt
Hmmm, Array-PAT-1.0.0.zip from CPAN is a ZIP, not a tar.gz, maybe that's
the problem? It certainly contains a Makefile.PL...

Of course, if any failing module is a prerequisite for other modules, it
causes a cascade of failures.

Practically everything in Apache:: fails due to some missing prerequisite.
Well, apache testing is something of a subject unto itself. I didn't look
at DBD/DBI, but I'll guess they're similar.

Just a few more from the 5.8 T section. Sorry, T is pretty large. It
includes Thread, Tie, and Tk - all large sub-groups. But there are a
suspicious number of Tk modules failed for "couldn't connect to display
:0".

This one failed for downrev prerequisite:
http://ppm.activestate.com/BuildSta...e-UTC-0.000.txt

I can see where it's difficult to keep up with versions, but the most
recent Time::UTC is dated late January this year. Math::BigRat's
changelog reports the required version was released in August 2002. From
various other revision problems it looks like AS is using modules from at
least a couple of years ago on their testers.

Here are some in the Tk area that failed with:
_X11TransSocketUNIXConnect: Can't connect: errno = 111
couldn't connect to display ":0"...
http://ppm.activestate.com/BuildSta...Control-1.0.txt
http://ppm.activestate.com/BuildSta...Dialog-0.10.txt
http://ppm.activestate.com/BuildSta...aphViz-0.13.txt
http://ppm.activestate.com/BuildSta...idEntry-1.0.txt

Thanks for taking an interest,

plugh

Sisyphus

2005-03-22, 3:56 am

Sisyphus

2005-03-23, 3:56 am

plugh wrote:
> On Wed, 16 Mar 2005 22:46:38 +0000, Sisyphus wrote:
>
>
>
>
> Well, I was looking more at the linux and solaris columns - I don't
> use DOS. I know AS is mostly Windows oriented, but they're including test
> results for other OSes too.
>

I'm not all that familiar with the non-Win32 side of AS ppm's, but there
are some things there that just don't look right so I sent a copy of
your (last) post to Jeff Griffiths at ActiveState. If he replies to me,
rather than to this NG, I'll forward his response here.

Cheers,
Rob

--
To reply by email u have to take out the u in kalinaubears.

plugh

2005-03-27, 8:55 am

On Fri, 18 Mar 2005 03:13:13 +0000, Sisyphus wrote:

> I'm not all that familiar with the non-Win32 side of AS ppm's, but there
> are some things there that just don't look right so I sent a copy of your
> (last) post to Jeff Griffiths at ActiveState. If he replies to me, rather
> than to this NG, I'll forward his response here.
>

Hello Sisyphus,

Any response yet from Active State on this? I haven't seen anything here
in clpm...

Regards,
plugh

Sisyphus

2005-03-29, 3:56 pm

--
To reply by email send to optusnet.com.au instead of nomail.afraid.org
"plugh" <plugh@hello--excise-this--six.com> wrote in message
news:pan.2005.03.27.08.34.07.356381@hello--excise-this--six.com...
> On Fri, 18 Mar 2005 03:13:13 +0000, Sisyphus wrote:
>
your[color=darkred]
rather[color=darkred]
>
> Hello Sisyphus,
>
> Any response yet from Active State on this? I haven't seen anything here
> in clpm...
>
> Regards,
> plugh

Yeah - sorry, I got sidetracked. Jeff got back to me, but only to say that
he was aware of those types of problems with the build status tables, and
would raise the matter with Jan Dubois (lead Perl devel at ActiveState). He
thought Jan would post to either me or the list, so I decided to wait a
coupla days and see what eventuated. Well .... nothing eventuated .... and
then I forgot about it ......

Jeff also provided the following link that searches *all* repositories, not
just ActiveState:
http://cpan.uwinnipeg.ca/htdocs/faqs/cpan-search.html . However, I'm not
sure whether that page accommodates non-Win32 ppm packages.

That's pretty much the gist of it - I've still got Jeff's reply on another
machine, but it's currently a little awkward to copy it.

Cheers,
Rob

plugh

2005-03-30, 3:56 pm

plugh

2005-03-30, 3:56 pm

On Wed, 30 Mar 2005 05:06:32 +1000, Sisyphus wrote:

plugh wrote:
>
> Yeah - sorry, I got sidetracked. Jeff got back to me, but only to say
> that
> he was aware of those types of problems with the build status tables,
> and would raise the matter with Jan Dubois (lead Perl devel at
> ActiveState).
> He thought Jan would post to either me or the list, so I decided to wait
> a coupla days and see what eventuated. Well .... nothing eventuated ....
> and then I forgot about it ......
>
> Jeff also provided the following link that searches *all* repositories,
> not just ActiveState:
> http://cpan.uwinnipeg.ca/htdocs/faqs/cpan-search.html . However, I'm not
> sure whether that page accommodates non-Win32 ppm packages.

> That's pretty much the gist of it - I've still got Jeff's reply on
> another machine, but it's currently a little awkward to copy it.

Sounds like Active State has no intention of fixing their problems.
Rather annoying given that without the perl developer community, they
wouldn't have much to offer. Yet they have no problems with falsely
labeling a large percentage of the perl developers as incompetent.

Turns the relationship from symbiotic towards parasitic in my opinion.

Well, thanks for trying, Sisyphus.

Regards,
plugh

plugh

2005-04-01, 3:57 pm

plugh

2005-04-06, 12:08 pm

Philippe M. Chiasson

2005-04-06, 8:55 pm

plugh wrote:
> On Wed, 30 Mar 2005 05:06:32 +1000, Sisyphus wrote:
> plugh wrote:
> [...]

Cutting to the meat of the issue.

I have recently joined the Perl crew at ActiveState, and all PPM issues have
fallen on my desk, so to speak ;-)

> Sounds like Active State has no intention of fixing their problems.

No, that's not quite the way I have been presented things. In the next couple
of months, a significant part of my time will be spent on
fixing/improving/redesigning major aspects og ActiveState's PPM building and
publishing infrastructure.

> Rather annoying given that without the perl developer community, they
> wouldn't have much to offer. Yet they have no problems with falsely
> labeling a large percentage of the perl developers as incompetent.

Yes, that is quite the case. At a glance, the ppm repository knows of about
116192 distinct build combinations (that's OS, Perl rev, CPAN module, etc)
and about 55383 of these have failed to build for one reason or another.
Automated building & testing of third-party software isn't an easy task
(especially when the software is CPAN as a whole).

I expect these numbers to significantly improve over time, as the automated
build system is improved. Please, bear with us and you should see the state
of things improve ;-) Of course, suggestions are always welcome!

> Turns the relationship from symbiotic towards parasitic in my opinion.

--------------------------------------------------------------------------------
Philippe M. Chiasson m/gozer\@(apache|cpan|ectoplasm)\.org/ GPG KeyID : 88C3A5A5
http://gozer.ectoplasm.org/ F9BF E0C2 480E 7680 1AE5 3631 CB32 A107 88C3A5A5

plugh

2005-04-10, 8:55 am

On Wed, 06 Apr 2005 20:41:34 +0000, Philippe M. Chiasson wrote:

> [ . . . ]
> I expect these numbers to significantly improve over time, as the
> automated build system is improved. Please, bear with us and you should
> see the state of things improve ;-) Of course, suggestions are always
> welcome!

Hello Philippe,

I am glad to hear that ActiveState is going to attempt to improve the
automated build/test system. I agree that it's not easy - you've got a
big job ahead of you. You may find some of the earlier articles in this
thread helpful to find a few starting points that might make significant
improvements without too much effort.

Hmmm, long-range, it might be useful to try to make some correlation of
your results with the CPAN tester results. I'm thinking it might be more
effective to put effort into modules that pass all or most CPAN tests
rather than wasting time on items that don't even work on the "lowest
common denominator". If you could add the CPAN results to your results
page, it might help people make more meaningful evaluations too.

Good luck with your new job,

plugh

plugh

2005-04-12, 8:56 pm

Sponsored Links