Show-stopping Bug in Module::Install and the Havoc it Created

2006-03-10 Thread Shlomi Fish
Hi all!

We had a discussion here about the pros and cons of using Module::Install with 
use inc::Module::Install in the Makefile.PL, and bundling it along. Well, 
now a showstopping bug happened in Module::Install and several modules that 
included the affected version suffered:

http://search.cpan.org/src/AUTRIJUS/Module-Install-0.59/Changes

http://rt.cpan.org/Public/Bug/Display.html?id=18020

http://rt.cpan.org/Public/Bug/Display.html?id=18090

Just for your information.

Regards,

Shlomi Fish

-
Shlomi Fish  [EMAIL PROTECTED]
Homepage:http://www.shlomifish.org/

95% of the programmers consider 95% of the code they did not write, in the
bottom 5%.


Re: Show-stopping Bug in Module::Install and the Havoc it Created

2006-03-10 Thread Tels
Moin,

On Friday 10 March 2006 12:26, Shlomi Fish wrote:
 Hi all!

 We had a discussion here about the pros and cons of using
 Module::Install with use inc::Module::Install in the Makefile.PL, and
 bundling it along. Well, now a showstopping bug happened in
 Module::Install and several modules that included the affected version
 suffered:

 http://search.cpan.org/src/AUTRIJUS/Module-Install-0.59/Changes

 http://rt.cpan.org/Public/Bug/Display.html?id=18020

 http://rt.cpan.org/Public/Bug/Display.html?id=18090

Just a litte bit of information: Module::Install 0.58 worked on my system, 
but didn't work on others. So the bug might not affect everyone - but it 
is bad and makes me re-consider using Module::Install only when 
necessary.

best wishes,

Tels

-- 
 Signed on Fri Mar 10 13:35:07 2006 with key 0x93B84C15.
 Visit my photo gallery at http://bloodgate.com/photos/
 PGP key on http://bloodgate.com/tels.asc or per email.

 What you have to have faith in is human beings being able to work
 together. -- Larry Sanger



pgpkCiLdeGo6l.pgp
Description: PGP signature


Re: Show-stopping Bug in Module::Install and the Havoc it Created

2006-03-10 Thread Yuval Kogman
On Fri, Mar 10, 2006 at 13:36:26 +0100, Tels wrote:

 So the bug might not affect everyone - but it is bad and makes me
 re-consider using Module::Install only when necessary.

Module::Install makes sense for applications and deployment and all
that For regular CPAN modules I think it's overkill.



-- 
 ()  Yuval Kogman [EMAIL PROTECTED] 0xEBD27418  perl hacker 
 /\  kung foo master: /me wields bonsai kittens: neeyah



pgpwVvPXpof4P.pgp
Description: PGP signature


Re: What is the Value of t/0-signature.t?

2006-03-10 Thread Audrey Tang
Yuval Kogman wrote:
 The problem with these is that Module::Signature fails when it
 should warn if the key is not present in the user's keyring.
 
 If this technical issue is solved then regardless of whether or not
 signature tests are useless (i think they are, but then again i also
 think i have some modules with Test::Distribution that checks
 that... so I am a hypocrit ;-) the usability of these tests will be
 good enough.

I think it should be like the standard Test::Pod's pod.t and only run
when an env var is set to true.

Patches... welcome to Module::Signature. :-)

Audrey



signature.asc
Description: OpenPGP digital signature


Re: Show-stopping Bug in Module::Install and the Havoc it Created

2006-03-10 Thread Adam Kennedy

Tels wrote:

Moin,

On Friday 10 March 2006 12:26, Shlomi Fish wrote:

Hi all!

We had a discussion here about the pros and cons of using
Module::Install with use inc::Module::Install in the Makefile.PL, and
bundling it along. Well, now a showstopping bug happened in
Module::Install and several modules that included the affected version
suffered:

http://search.cpan.org/src/AUTRIJUS/Module-Install-0.59/Changes

http://rt.cpan.org/Public/Bug/Display.html?id=18020

http://rt.cpan.org/Public/Bug/Display.html?id=18090


Just a litte bit of information: Module::Install 0.58 worked on my system, 
but didn't work on others. So the bug might not affect everyone - but it 
is bad and makes me re-consider using Module::Install only when 
necessary.


Actually, that it worked on your system was intended behaviour. The bug 
that was introduced meant that installation failed only for authors 
(that is, that had Module::Install installed on their local system) and 
only if their installed version differed from the one bundled.


Yes, I fucked up. But it's hardly a showstopper.

It only caused problems for authors. Although that was a little 
unfortunate given that Catalyst users seem to now all be authors by 
definition, it meant normal end-users shouldn't have been effected (I 
think).


Module::Install was broken for less than a week, and only for authors.

Any author who upgrades to the just uploaded 0.60 will be fixed now.

As mentioned before, the design of Module::Install tends to favour the 
end-user over authors, because authors are a small subset of all users 
and are also more capable of locating and fixing problems.


I apologise for any inconvenience over the last week.

But I think this demonstrates the strength of the design of MI, rather 
than is being a negative.


In the cases MI does have a major problem, it can be located fairly 
quickly, fixed fairly quickly, and once fixed leaves minimal residue 
around to bother us down the track. There's no back-compatibility issues 
to deal with once we have fixed the problem.


And yes, I agree with TELS. MI is NOT 1.00 yet, and so various small 
problem may occur from time to time as we go through the refactoring of 
it to add VMS support and all the other things that take it from being a 
cool installer into a highly robust system that works everywhere.


But then the ONLY things that are production grade and completely stable 
are EU:MM and CPAN.pm.


And if your module is simple enough that using raw EU:MM is 
satisfactory, I recommend people stick with it for now. I'm not going to 
hype MI just because it's new and shiny.


And if you do need a feature in MI we welcome the additional user, and 
we'll help you in any way we can. I hope the responsiveness to your bugs 
for example have been appreciated (btw, the security information leak is 
fixed now in the newly uploaded 0.60).


Module::Install, Module::Build and CPANPLUS are all not yet ready for 
prime time yet, and so if absolute back-compatibility is essential, 
stick to EUMM for now.


And I make no apologies for preferencing millions of end users over 
thousands of authors, even though I happen to talk to authors more than 
to users.


All authors are recommended to upgrade to Module::Install 0.60 as soon 
as possible. Modules released with 0.58 should be upgraded to at least 
0.59 or higher. All modules released with 0.59 do not need to be upgraded.


Thanks for your patience, and I hope you keep those bug reports rolling in.

Adam K