Re: G-O-I under MinGW

[LRN <lrn1986 gmail com>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 16.04.2014 19:48, Martin Schlemmer wrote:
> On 2014/04/16 at 02:57 PM, LRN wrote:
> > On 16.04.2014 16:43, Martin Schlemmer wrote:
> > > At some point LRN wrote:
> > > > gobject-introspection might fail to generate stuff (failure
> > > > at shutil.rmtree() in gdumpparser.py), especially on slow
> > > > machines. Re-run the build from the last step. No new
> > > > insights into this bug.
> > >
> > > Attached is an old patch (I used Dieterv's G-O-I branch) that
> > > fixes this mostly for me.
> > >
> > > From what I could determine at the time, either the Anti-Virus
> > > or some other culprit still had the .exe locked, so the rm
> > > failed the first time. Waiting (might make it longer for slower
> > > machines) on failure seems to fix it.
> >
> > I already have something similar[1] applied.
>
> On the G-O-I subject, and since you mentioned Webkit, there is
> another issue with Webkit and its generated DOM bindings when it
> comes to G-O-I.
>
> At least here, the generated code have Win line endings, and this
> breaks the scanner. This happens in two ways:
>
> - scannerlexer.l happily adds the CR's, but parse_comment_block()
> from annotationparser.py splits on LF, so the regex's scanning for
> comment blocks fails and results in no comment blocks being
> attached to functions, etc., causing introspectable="0" all over
> the place
>
> - little less an issue, but because gcc's output in
> sourcescanner.py via a pipe, somewhere MSYS's line ending code get
> things wrong, and you end up with <CR><CR><LF> instead of just
> <CR><LF>. The only impact here is missing:
>
> <doc xml:whitespace="preserve">...</doc>
>
> bits, but I have not really tried to track the P.O.F. down, I just
> avoided the pipe.
>
> What I did for these was: - strip CR's in scannerlexer.l (if
> followed by a LF) - could probably fix it in annotationparser.py,
> but this seemed easier - use gcc's -o argument instead of piping
> the output
>
> The patch is against 1.36.0 is tested, the one against 1.38.0 is
> not with Webkit at least. If I get time I will update to the latest
> sources and see if anything new breaks.

Patch looks sane (that said, i expect GI devs to claim that fixing
annotationparser.py is the Right Thing (TM) to do). I'll try it out
when i get to updating GI package.


- -- 
O< ascii ribbon - stop html email! - www.asciiribbon.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (MingW32)

iQEcBAEBAgAGBQJTTrGuAAoJEOs4Jb6SI2CwWroH/RIqQdggrXA5qgALhS+/DYIk
7qEJUqKSOVdUw+zv2yq83xlVTZuiCFzaKIvUHfukMG/DK4iY2kAturJX6Li0FU0h
podi9p+Q1oKXQiP+IeqiEHV8PhxWaTcU3GMWHH7MsE6RC7thKJ2aJtHmIAu3HnPk
tblr6Msf/klXax8z4JvaK8Z0AXhnCTn2qcrol1LCye/W9Cr/btMx4GoXJyBgq62A
nSSff+42nDx2rW0BxHbnkLiJkAO73GIKUGMnCTiFrdlUU+5ArDziIbB+akR2ZObm
zu/+MafV26+MkyT6UZOqLXFh73lcWBZfzPznh3OP90IKT9IKou3EhBi4ejlo1xo=
=hLqL
-----END PGP SIGNATURE-----