Eric Wilhelm | 5 Dec 22:54 2007
Picon

huh? Fwd: FAIL Math-Round-Var-0.03 i686-linux 2.4.20-8smp

Something to do with the version sub and/or the compat layer?

----------  Forwarded Message:  ----------

Subject: FAIL Math-Round-Var-0.03 i686-linux 2.4.20-8smp
Date: Wednesday 05 December 2007 12:01
From: alexchorny <at> gmail.com
To: cpan-testers <at> perl.org
Cc: EWILHELM <at> cpan.org

This distribution has been tested as part of the cpan-testers
effort to test as many new uploads to CPAN as possible.  See
http://testers.cpan.org/

Please cc any replies to cpan-testers <at> perl.org to keep other
test volunteers informed and to prevent any duplicate effort.

--
Dear Eric Wilhelm,

This is a computer-generated report for Math-Round-Var-0.03
on perl-5.00505, created automatically by CPAN-Reporter-1.04
and sent to the CPAN Testers mailing list.

If you have received this email directly, it is because the person
 testing your distribution chose to send a copy to your CPAN email
 address; there may be a delay before the official report is received
 and processed by CPAN Testers.

Thank you for uploading your work to CPAN.  However, it appears that
(Continue reading)

Dave Rolsky | 5 Dec 23:05 2007

Re: huh? Fwd: FAIL Math-Round-Var-0.03 i686-linux 2.4.20-8smp

On Wed, 5 Dec 2007, Eric Wilhelm wrote:

> Something to do with the version sub and/or the compat layer?

I got a similar report from this Alex for a module of mine on Perl 5.005, 
and I also used the passthrough Makefile.PL.

However, looking at where it's failing my guess is that the code just 
doesn't work on 5.005, because of the use of "our".

-dave

/*===================================================
VegGuide.Org                        www.BookIRead.com
Your guide to all that's veg.       My book blog
===================================================*/

David Golden | 5 Dec 23:09 2007
Picon

Re: huh? Fwd: FAIL Math-Round-Var-0.03 i686-linux 2.4.20-8smp

Your modules probably define their version with "our $VERSION" -- that
blows up on 5.005 when EU::MM or M::B try to evaluate the line they
parse out of your code.  If you put "use 5.006" at the top of your
Makefile.PL or Build.PL, you'll be much friendlier to anyone who tries
to run your code on 5.005.  On the other hand, Schwern declared 5.005
dead so you may not care.

However, Slaven's report is also from an older version of
CPAN::Reporter  -- in 1.05 I added an explicit check to detect this
situation and issue an NA grade instead of a FAIL.

David

On Dec 5, 2007 5:05 PM, Dave Rolsky <autarch <at> urth.org> wrote:
> On Wed, 5 Dec 2007, Eric Wilhelm wrote:
>
> > Something to do with the version sub and/or the compat layer?
>
> I got a similar report from this Alex for a module of mine on Perl 5.005,
> and I also used the passthrough Makefile.PL.
>
> However, looking at where it's failing my guess is that the code just
> doesn't work on 5.005, because of the use of "our".
>
>
> -dave
>
> /*===================================================
> VegGuide.Org                        www.BookIRead.com
> Your guide to all that's veg.       My book blog
(Continue reading)

Eric Wilhelm | 5 Dec 23:41 2007
Picon

Re: huh? Fwd: FAIL Math-Round-Var-0.03 i686-linux 2.4.20-8smp

# from Dave Rolsky
# on Wednesday 05 December 2007 14:05:

>However, looking at where it's failing my guess is that the code just
>doesn't work on 5.005, because of the use of "our".

Oh, right.  Missed the 5.005 bit.  Thanks.

--Eric
--

-- 
"It ain't those parts of the Bible that I can't understand that
bother me, it's the parts that I do understand."
--Mark Twain
---------------------------------------------------
    http://scratchcomputing.com
---------------------------------------------------

Thomas Wittek | 11 Dec 14:41 2007
Picon

requires vs. build_requires in CPAN

Hi!

Recently I uploaded a Module [1] to CPAN that uses M::B for the build
process.
Now, when you try to install it CPAN fails at the test cases as some
dependencies are missing.

Do I have to put all dependencies that are needed for "./Build test"
into 'build_requires'?
If yes, then 'requires' would be completely empty, as I have at least
one test for each module.

Or do I have to do anything else to get CPAN to recognize my
dependencies correctly?

Thank you!

[1] http://search.cpan.org/~twittek/Konstrukt-0.5-beta8/

Adriano Ferreira | 11 Dec 18:08 2007
Picon

Re: requires vs. build_requires in CPAN

On Dec 11, 2007 11:41 AM, Thomas Wittek <mail <at> gedankenkonstrukt.de> wrote:
> Hi!
>
> Recently I uploaded a Module [1] to CPAN that uses M::B for the build
> process.
> Now, when you try to install it CPAN fails at the test cases as some
> dependencies are missing.
>
> Do I have to put all dependencies that are needed for "./Build test"
> into 'build_requires'?
> If yes, then 'requires' would be completely empty, as I have at least
> one test for each module.
>
> Or do I have to do anything else to get CPAN to recognize my
> dependencies correctly?

By a quick inspection, I think your sin was to release a distribution
with only Build.PL and no passthru Makefile.PL. Some CPAN testers
(among them, chris) do not install in their smoke servers the part of
CPANPLUS necessary for running Build.PL alone and that leads to ugly
failures. The default behavior (nonsense to me) is to try to generate
automatically a Makefile.PL which misses all dependencies and have all
but an inferred dist name and a version. That is a safe road to FAIL
in most of the cases.

I was plagued by that too, but only with an Acme module which I insist
to leave that way. If the testers are using the CPAN shell, test
proceeds smoothly for these distributions.

So what you need is just use the M::B option "create_makefile_pl" as
(Continue reading)

Michael G Schwern | 12 Dec 01:27 2007
Face
Picon

Re: requires vs. build_requires in CPAN

Adriano Ferreira wrote:
> By a quick inspection, I think your sin was to release a distribution
> with only Build.PL and no passthru Makefile.PL. Some CPAN testers
> (among them, chris) do not install in their smoke servers the part of
> CPANPLUS necessary for running Build.PL alone and that leads to ugly
> failures. The default behavior (nonsense to me) is to try to generate
> automatically a Makefile.PL which misses all dependencies and have all
> but an inferred dist name and a version. That is a safe road to FAIL
> in most of the cases.

FWIW I talked with Jos about it and I'm working up a patch to make CPANPLUS Do
The Right Thing, which is to install Module::Build as an implied prereq.

> PS. You may want to add "configure_requires => 'Module::Build' " as
> well. But most CPANPLUS-based toolchain does not honor yet (or so I
> think).

I worked with Jos and submitted a patch for that.
http://rt.cpan.org/Public/Bug/Display.html?id=29676

Still needs a little work, like tests and docs and stuff, but it looks like
the CPANPLUS vs Module::Build fight will be over soon.

--

-- 
Hating the web since 1994.

Graham TerMarsch | 12 Dec 01:57 2007

Re: requires vs. build_requires in CPAN

On Tuesday 11 December 2007 4:27 pm, Michael G Schwern wrote:
> FWIW I talked with Jos about it and I'm working up a patch to make CPANPLUS
> Do The Right Thing, which is to install Module::Build as an implied prereq.
>
> > PS. You may want to add "configure_requires => 'Module::Build' " as
> > well. But most CPANPLUS-based toolchain does not honor yet (or so I
> > think).
>
> I worked with Jos and submitted a patch for that.
> http://rt.cpan.org/Public/Bug/Display.html?id=29676
>
> Still needs a little work, like tests and docs and stuff, but it looks like
> the CPANPLUS vs Module::Build fight will be over soon.

Thank you. :)

I'll be so happy when I don't have to jump through hoops to get things to fail 
gracefully when CPANPLUS already knows that it doesn't have everything it 
needs to do a successful build.

-----

The last CPANPLUS/M::B thing that'd still irk me is M::B runs in-process with 
CPANPLUS.  I've already had this bite me hard with two separate scenarios:

1) CPANPLUS testers aren't able to send any useful pass/failure output to let 
me know what happened.

2) If you use "cpanp" to try to install more than one package, you get 
namespace collisions (e.g. "MY::Build" clashes).
(Continue reading)

Ken Williams | 12 Dec 04:49 2007
Picon

Re: requires vs. build_requires in CPAN


On Dec 11, 2007, at 7:41 AM, Thomas Wittek wrote:

> Hi!
>
> Recently I uploaded a Module [1] to CPAN that uses M::B for the build
> process.
> Now, when you try to install it CPAN fails at the test cases as some
> dependencies are missing.
>
> Do I have to put all dependencies that are needed for "./Build test"
> into 'build_requires'?
> If yes, then 'requires' would be completely empty, as I have at least
> one test for each module.

I think other people helped figure out the actual problem, but for  
the record here's the answer to your original question.

'build_requires' is the extra stuff that is *only* needed during  
building & testing.  It's understood that all of 'requires' will also  
be used during building & testing, since the code should in general  
be fully exercised in the tests.

  -Ken

Adam Kennedy | 12 Dec 05:07 2007
Picon

Re: requires vs. build_requires in CPAN

The others have answered your question.

But I'll note that the pass-through Makefile.PL is required to allow for
installation on default installs from any Perl distribution 5.8.8 or older
(due to a long-standing circular problem that prevents auto-upgrading).

5.10.0 will support the current Module::Build out of the box, and we're
getting close to having full configure_requires support across the whole
toolchain (with formal support being ticked off in 5.10.1).

The pass-through will continue to be required for a few years yet to handle
the 90% of the userbase that don't immediately upgrade, but at some point
we'll do a cutover and stop supporting anyone without configure_requires
support.

Adam K

On 12/12/2007, Thomas Wittek <mail <at> gedankenkonstrukt.de> wrote:
>
> Hi!
>
> Recently I uploaded a Module [1] to CPAN that uses M::B for the build
> process.
> Now, when you try to install it CPAN fails at the test cases as some
> dependencies are missing.
>
> Do I have to put all dependencies that are needed for "./Build test"
> into 'build_requires'?
> If yes, then 'requires' would be completely empty, as I have at least
> one test for each module.
(Continue reading)


Gmane