Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-04-23

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:01 jeffreykegler joined #marpa
02:55 ronsavage joined #marpa
03:57 jeffreykegler1 joined #marpa
07:19 ronsavage joined #marpa
15:30 jeffreykegler joined #marpa
18:33 jdurand joined #marpa
18:33 jdurand Jeffrey, followup to https://github.com/jeffreykegler/Marpa--R2/pull/135 :
18:34 jdurand you want to get rid of Build_Me.pm - about Build.PL mechanism as well ?
18:44 jeffreykegler Module::Build is being removed from the core -- it's semi-deprecated.
18:44 jeffreykegler Frankly, it's a bit of a disaster.
18:46 jeffreykegler It's an object-oriented experimentation gone seriously wrong.
18:49 jeffreykegler The idea seems to have been that it provided a lot of building blocks, as well as an overall framework.  But what it meant was that, even deep inside the code, it was impossible to tell whether a behavior was something depended on on the outside.  So it quickly became impossible to change the code and risky to maintain it.
18:52 jeffreykegler Getting rid of Build_Me.pm is not a priority, but down the road I may wind up forced to redo things in ExtUtils::MakeMaker.  So I'm currently making sure that, when and if the crisis comes, the move will be as easy as possible.
18:54 jeffreykegler One thing that means is that, while functionality of existing tasks may be extended, new tasks are not added to Build.PL and Build_Me.pm.
19:05 jdurand Ok - may I say that when I moved to Dist::Zilla, took me one week to find and understand its subtilities, but it rocks
19:07 jdurand I am currently writing this stand-alone script as of #135 - then per 99% of do_libmarpa could be replaced by a call to it, but this routine can stay as it is also
19:07 jeffreykegler Re Dist::Zilla:  Abraham Lincoln used to point out that, once a cat has been burned, he won't even sit on a cold stove.
19:08 jeffreykegler I've have good things about Dist::Zilla, but that was the case with Module::Build -- and I don't think Dist::Zilla is in the core, is it?
19:09 jeffreykegler Not that being in the core proved much help in the case of Module::Build :-)
19:10 jdurand It is not - and has been attacked one month ago on blogs.perl.org in fact - by people that simply did not understood how to use it - Dist:Zilla is built on top on Moose - and that is a stable relationship
19:10 jeffreykegler jdurand: re http://irclog.perlgeek.de/marpa/2014-04-23#i_8629062 -- I didn't fully understand this.
19:11 jdurand it is IMHO a remarquable OO package, one of the bests - controlled by R.Signes which is another good point
19:11 jeffreykegler My opinion now is that installation and distribution management should be handled by separate packages -- intertwining them is a bad idea.
19:12 jeffreykegler An installer is expected to work remotely, but only has a few tasks it needs to do -- configure, build, test, & install.
19:12 jdurand Dist::Zilla offers this. you need to build the package. But the package itself will not depend on it.
19:13 jeffreykegler Distribution management you want lots of facilities for, but it's done in the developer environment and you can be aggressive about adding dependencies.
19:14 jeffreykegler What does a Dist::Zilla package use for installation?  EU::MM?
19:14 jdurand Re: http://irclog.perlgeek.de/marpa/2014-04-23#i_8629062  - with this stand-alone script do_libmarpa will simply become: call the script, then build. But that is not obligatory. That's why I'll do two separate pulls
19:14 jdurand As you like: EU;MM or MB
19:14 jdurand Default is EU:MM
19:14 jeffreykegler Guess which I'd choose. :-)
19:20 jeffreykegler Another thing to bear in mind here is that building the shared library needs to be completely independent of building Marpa::R2.  Things get confused a bit, because apparently it's ok to assume Perl on Windows, so Perl can be used to build the shared library, but even so, the shared library should be for someone who might even be Perl-adverse.
19:20 jeffreykegler Whereas, of course, the deeply Perl-adverse are not going to write applications build on Marpa::R2
19:22 jdurand Yes, this is a difficulty with current packaging. libmarpa is too much bundled with perl because it is in a CPAN package, simply
19:23 jdurand script finished - testing -;
19:25 jdurand Damned - works
19:27 jdurand And testing it I realize there is a bug in Build_Me.pm
19:33 jdurand Btw compiling with -DMARPA_DEBUG=1 exhibits a lot of warnings - sprintf undeclared
19:34 jeffreykegler Is this still about building shared libraries?
19:35 jdurand Yes, a stand-alone script that generates libmarpa_build with a config.h
19:36 jeffreykegler So Build_Me.pm is not involved?
19:37 jeffreykegler Also, when compiling with -DMARPA_DEBUG=1 under configure, I haven't worried about warnings.  I leave in few in under Linux to remind me that I've compiled in debug mode.
19:38 jeffreykegler So I don't necessarily regard warnings with  -DMARPA_DEBUG=1 as problems.  Compiling for production is another matter.
19:40 jdurand Yes, Build_Me.pm is not involved. I am going to pull it.
19:40 jeffreykegler On the other hand, if there are so many warnings that it makes it hard to work with the Marpa code in a environment that's important, it may be useful to silence some proportion of them.
19:40 jdurand Take it as experimental for the moment. This will be used only by the Win32 Makefile. Better not to touch Build_Me.pm beast ater call.
19:40 jdurand "after all"
19:41 jeffreykegler Some of debug-only code uses sprintf.
19:42 jdurand No pb. forget about my debug mode comments.
19:43 jdurand Pulling
19:46 jdurand Gree how can I pull only the latest one
19:58 jdurand Done -; somehow github can confused by me doing git reset 'HEAD^' --hard three times, pushing, and doing a new commit/push!... the latest commit got assigned to #135 and I do not know if this is normal. So I closed #135 definitely and now you have #137
20:02 jeffreykegler I'll look at it after lunch.  Thanks!
20:09 yxhuvud It should be pretty simple to check that everything is alright by cloning the repo, or by doing git reset --hard origin/master. (the latter may be best done in a new branch to avoid confusion, if you are not familiar enough with git).
20:14 jdurand yxhuvud - yep thnx a log - I realise I could have done a git reset playing with upstream branch, perhaps... I am not an experienced user with git, definitely!
20:14 jdurand "thnx a lot"
20:16 yxhuvud one potentially confusing part when doing reset locally is that it doesn't propagate to master when pushing (unless you use a certain flag for that, but it is generally considered bad style unless the shit has hit the fan).
20:16 jdurand you re right. Had to do git push -f
20:17 jdurand Sign that something was not clean - anyway now it is ok -;
20:17 * yxhuvud spent my first years with git doing every mistake under the sun. gives a good knowledge of how to make things work but it was not a very efficiently spent time
20:18 jdurand Have you played a bit with Marpa since you re on this channel
20:18 yxhuvud I'm implementing my own :D
20:19 yxhuvud it can recognize a grammar but is not far enough along yet.
20:20 jdurand Jeffrey, there is a spy on the channel! laughs -; No pb, you're welcome and any suggestion of any kind will be well received
20:20 yxhuvud your work extracting the lib is very interesting
20:20 jdurand Ah, thanks -;
20:20 yxhuvud cpan dependency is a gamekiller for me.
20:23 jdurand Yes, it makes things more difficult in case of libmarpa - the best way would be to have a separate distribution, in fact what I have done for debian.
20:24 jdurand Perhaps this will be one of the issues adressed with next major version - anyway, every design issue has to be acked by Jeffrey
20:25 yxhuvud the beauty of writing my own - noone to stop my stupid ideas!
20:27 jdurand I'll be happy to help if your project stands up on the web - but I am fundamentally a technician in writing code - not a mathematician alike Jeffrey
20:27 jdurand Ok, time for me to go sleeping - looking forward news from you -;
20:27 jdurand AFK
20:27 yxhuvud it will, eventually when I have more parts ready (and cleaned up the code organization a bit
20:28 jdurand -;

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary