[DBIx-Class-Devel] Busted Releasing on Windows

[Peter Rabbitson]

On Thu, Dec 13, 2012 at 10:44:59AM -0600, fREW Schmidt wrote:
> haarg noticed some weird stuff with the build proc causing
> SineSwiper's strange build:
> 
>    < haarg> that selfinject-pod thing in dbicadmin looks rather goofy
>    < haarg> but aside from that i'm not sure how to make that work with dmake
>    < haarg> in addition to the exec call looking very broken on windows
>    < frew> haarg: where is this, I'm looking and can't find it
>    < haarg> http://git.shadowcat.co.uk/gitweb/gitweb.cgi?p=dbsrgits/DBIx-Class.git;a=blob;f=script/dbicadmin;h=58ef4c823e1634f98cad750fbe66a4281bca5f02;hb=master#l71
>    < haarg> used here 
>    adowcat.co.uk/gitweb/gitweb.cgi?p=dbsrgits/DBIx-Class.git;a=blob;f=maint/Makefile.PL.inc/52_inject_dbicadmin_pod.pl;h=e9f098060d6b5e1dc6943941397366cab5e0202a;hb=master
>    < frew> oh goodness
>    < haarg> dmake doesn't set MAKELEVEL
>    < frew> so a stupid fix would be to just remove the die entirely
>    < frew> and that would work?
>    < haarg> possibly
>    < haarg> that exec call still won't work though
>    < frew> why?  won't that be called by perl itself?  not the Makefile?
>    < haarg> honestly i'm tempted to say "never do releases from windows"
>    < frew> (ie Makefule.PL)
>    < frew> oooh
>    < frew> yeah
>    < frew> that's legit
>    < frew> if you release on windows your perms will be all screwed up anyway
>    < haarg> well, actually that's at duild time, not release time
>    < haarg> build
>    < frew> (in my experience)
>    < haarg> or no
>    < haarg> bleh sorry
>    < haarg> that's release time
>    < haarg> and yeah the permissions are all screwed
>    < haarg> and possible the M::I stuff is busted too
>    < frew> so maybe a better soln would be at the beginning die "use a real OS to release DBIC" if $^O =~ m/^(?:Win32|cygwin)$/i
>    < haarg> SineSwiper's dev release had a number of problems like that
>    < haarg> possibly
>    < frew> I'm gonna just mail this to the list and see what riba thinks, since he made this weird M::I framework
> 
> So basically I posit that we should bail builds done on windows due to
> a menagerie of fun and probably not worth fixing issues.

In general I would say "yes, packaging on win32 is batshit insane". However
I think EUMM actually offers me the tools to do it properly. Today was
very fucked up so no code yet, but it's in the pipe.

As far as thre crlf issue - this is why make dist runs a prove -l xt.

Cheers