Re: [PATCH] I must be /mad/

On May 5, 2008, at 12:56 PM, Gerard Goossen wrote:
> First some more general note about you working on MAD.
> 1. Great.
> 2. I have many more MAD changes take a look at the bleadgerard  
> branch at
> git://dev.tty.nl/ the commits have to be tidied up before they are
> ready to be applied, but there are a lot of MAD fixes and changes  
> there.

Excellent.

>
> Moving things to lib seems a good idea, to me, but shouldn't we go a
> little further and move MAD to ext/ like what the plan is for dual- 
> lived modules.

Nick and I talked this through on #p5p. The advice I got was "for now,  
move everything to lib/. ext/ can come later if it's deemed to make  
sense"

>
> Or maybe even better, since it already has a dependency on  
> XML::Parser, to
> remove it from the core and make it a CPAN module.

That strikes me as dangerous in the absence of an alternate  
comprehensive MAD test suite.  Right now, the only tests MAD has are  
roundtripping the rest of the test suite.

Removing all the tests for MAD from the core seems to be the surest  
way to kill it entirely.

Larry mentioned possibly moving MAD's output away from XML.  
Alternatively, the XML is also well-enough-defined that we could  
likely replace XML::Parser with something trivial. In the meantime,  
having it _possible_ to run the tests with less heartache seems like a  
win.

>
>
> As a technical note about your patch, please make the move of the  
> files
> separate from the changes, i.e. first instructions to move files and  
> then a
> patch of the changes you made to the moved files.

Pretty much all of the changes to the files were renaming the classes  
as part of the move. Doing it as two separate operations would result  
in a completely-broken middle step. rgs: Is that interim step  
something you'd need to be able to apply this?

-j

>
>
> On Sat, May 03, 2008 at 10:44:31PM -0400, Jesse Vincent wrote:
>> I spent four hours today getting my hair turned bright purple for
>> YAPC::Asia. So I had a bit of downtime. I spent it tidying up MAD for
>> 5.10, resulting in my first and second non-documentation core  
>> patches.
>>
>> Briefly:
>>
>> Based on a brief chat with Nicholas, I moved MAD's libraries into  
>> lib/.
>>
>> Based on a brief chat with Larry, I moved MAD's libraries into the  
>> MAD::
>> namespace.
>>
>> I removed the 'p55' program, as it was fully replaced by the tests  
>> that
>> Gerrard wrote. (It was just a test harness for basic command-line
>> testing).
>>
>> I created a 'utils/nomad' utility script to bring back  
>> functionality we
>> lost when Gerrard moved the original script into a library  
>> (Nomad.pm, now
>> MAD::Nomad.pm)
>>
>> I cleaned up Gerrard's test script such that it can run as part of  
>> the
>> test suite, with the caveat that Nomad still depends on XML::Parser,
>> which isn't in core. The tests now skip if you don't magically have  
>> an
>> XML::Parser or if perl was not built with -Dmad=y
>>
>> Only two of MAD's non-todoed tests are failing once you apply the  
>> patch I
>> sent to dump.c earlier today. (Before that it was 30 or so).
>>
>> There's still some bitrot, but this goes a ways toward getting things
>> back to usable and testable.
>>
>> Best,
>> Jesse
>>
>>
>>
>
>
>>
>
>
>
>
> -- 
> Gerard Goossen
> TTY Internet Solutions
> http://dev.tty.nl/
>
>

• [PATCH] I must be /mad/ by Jesse Vincent
• Re: [PATCH] I must be /mad/ by Gerard Goossen
• Re: [PATCH] I must be /mad/ by Jesse Vincent