Received: by 10.58.202.234 with SMTP id kl10mr8506550vec.33.1352875818981;
        Tue, 13 Nov 2012 22:50:18 -0800 (PST)
X-BeenThere: marpa-parser@googlegroups.com
Received: by 10.52.67.52 with SMTP id k20ls126373vdt.1.gmail; Tue, 13 Nov 2012
 22:50:18 -0800 (PST)
Received: by 10.52.20.51 with SMTP id k19mr73015vde.4.1352875818682;
        Tue, 13 Nov 2012 22:50:18 -0800 (PST)
Date: Tue, 13 Nov 2012 22:50:17 -0800 (PST)
From: Durand Jean-Damien <jeandamiendur...@gmail.com>
To: marpa-parser@googlegroups.com
Message-Id: <8f628be5-8799-4a90-a12c-c3e3f774beca@googlegroups.com>
In-Reply-To: <50A1BCB2.7030307@jeffreykegler.com>
References: <50A1BCB2.7030307@jeffreykegler.com>
Subject: Re: Windows-capable version (?) or Marpa::R2 on CPAN
MIME-Version: 1.0
Content-Type: multipart/mixed; 
	boundary="----=_Part_493_9425597.1352875817127"

------=_Part_493_9425597.1352875817127
Content-Type: multipart/alternative; 
	boundary="----=_Part_494_5539582.1352875817127"

------=_Part_494_5539582.1352875817127
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

Jeffrey,
FYI the RT ticket #81064<https://rt.cpan.org/Public/Bug/Display.html?id=3D8=
1064>on Config::AutoConf being solved, you might want to give a recommended=
=20
dependency on version 0.20 on Config::AutoConf, if that's possible and=20
replace the awful line in Build_Me.pm that contained stdint.h by a single=
=20
line $ac->check_default_headers(). Later, at least let's have a full=20
working release on Windows at least one -;
Cheers, JD.

Le mardi 13 novembre 2012 04:21:31 UTC+1, Jeffrey Kegler a =C3=A9crit :
>
> I took Jean-Damien's Windows port upstream, and it is now on CPAN as=20
> Marpa::R2 2.025_000.  Cpantesters troubles continue (though there is=20
> progress) and I'd appreciate seeing test results from list members.  I=20
> made some serious changes to the build procedure on POSIX, so there may=
=20
> be regression issues on UNIX and Linux.=20
>
> Notes about the port:=20
>
> 1.)  I realized there was no good reason for me to be untar'ing files=20
> during the build and so eliminated all use of tar files.  This removes=20
> the need for Archive::Tar.  It does mean a considerable rearrangement of=
=20
> the way the build works, even on POSIX/UNIX/Linux.=20
>
> 2.)  On Windows, Build.PL tries to load Config::Autoconf and die's if it=
=20
> cannot.  Config::Autoconf is not actually needed until Build is run, but=
=20
> the way cpantesters works, it is best to signal a failure from Build.PL,=
=20
> before generating a Build file.  Since Config::Autoconf is only=20
> recommended, this is a somewhat annoying way to tell a Windows user that=
=20
> he needs Config::Autoconf, but this seems to be the least bad option.=20
>
> 3.)  I kept Jean-Damien's autoconf logic almost 100% intact.  One change=
=20
> was the elimination of explicit checking of a couple of the open calls=20
> -- the convention in my build is that Fatal catches open() errors.  =20
> Arguably, in substituting a catch-all for a hand-crafted error message,=
=20
> I replaced Jean-Damien's better code with my worse code.  But it's best=
=20
> to have consistency in error handling.=20
>
> 4.) The environment variable to force use of Jean-Damien's install path,=
=20
> even on POSIX, is MARPA_USE_PERL_AUTOCONF.  It's an interesting=20
> alternative.=20
>
> I tested as best I can without access to Windows.  It won't surprise me=
=20
> if this takes several iterations to get right.=20
>
> -- jeffrey kegler=20
>

------=_Part_494_5539582.1352875817127
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: quoted-printable

Jeffrey,<br>FYI the RT ticket <a href=3D"https://rt.cpan.org/Public/Bug/Dis=
play.html?id=3D81064">#81064</a> on Config::AutoConf being solved, you migh=
t want to give a recommended dependency on version 0.20 on Config::AutoConf=
, if that's possible and replace the awful line in Build_Me.pm that contain=
ed stdint.h by a single line $ac-&gt;check_default_headers(). Later, at lea=
st let's have a full working release on Windows at least one -;<br>Cheers, =
JD.<br><br>Le mardi 13 novembre 2012 04:21:31 UTC+1, Jeffrey Kegler a =C3=
=A9crit&nbsp;:<blockquote class=3D"gmail_quote" style=3D"margin: 0;margin-l=
eft: 0.8ex;border-left: 1px #ccc solid;padding-left: 1ex;">I took Jean-Dami=
en's Windows port upstream, and it is now on CPAN as=20
<br>Marpa::R2 2.025_000. &nbsp;Cpantesters troubles continue (though there =
is=20
<br>progress) and I'd appreciate seeing test results from list members. &nb=
sp;I=20
<br>made some serious changes to the build procedure on POSIX, so there may=
=20
<br>be regression issues on UNIX and Linux.
<br>
<br>Notes about the port:
<br>
<br>1.) &nbsp;I realized there was no good reason for me to be untar'ing fi=
les=20
<br>during the build and so eliminated all use of tar files. &nbsp;This rem=
oves=20
<br>the need for Archive::Tar. &nbsp;It does mean a considerable rearrangem=
ent of=20
<br>the way the build works, even on POSIX/UNIX/Linux.
<br>
<br>2.) &nbsp;On Windows, Build.PL tries to load Config::Autoconf and die's=
 if it=20
<br>cannot. &nbsp;Config::Autoconf is not actually needed until Build is ru=
n, but=20
<br>the way cpantesters works, it is best to signal a failure from Build.PL=
,=20
<br>before generating a Build file. &nbsp;Since Config::Autoconf is only=20
<br>recommended, this is a somewhat annoying way to tell a Windows user tha=
t=20
<br>he needs Config::Autoconf, but this seems to be the least bad option.
<br>
<br>3.) &nbsp;I kept Jean-Damien's autoconf logic almost 100% intact. &nbsp=
;One change=20
<br>was the elimination of explicit checking of a couple of the open calls=
=20
<br>-- the convention in my build is that Fatal catches open() errors. &nbs=
p;
<br>Arguably, in substituting a catch-all for a hand-crafted error message,=
=20
<br>I replaced Jean-Damien's better code with my worse code. &nbsp;But it's=
 best=20
<br>to have consistency in error handling.
<br>
<br>4.) The environment variable to force use of Jean-Damien's install path=
,=20
<br>even on POSIX, is MARPA_USE_PERL_AUTOCONF. &nbsp;It's an interesting al=
ternative.
<br>
<br>I tested as best I can without access to Windows. &nbsp;It won't surpri=
se me=20
<br>if this takes several iterations to get right.
<br>
<br>-- jeffrey kegler
<br></blockquote>
------=_Part_494_5539582.1352875817127--

------=_Part_493_9425597.1352875817127--