svn commit: r303626 - head/www/xpi-firebug

[Alex Kozlov]

On Tue, Sep 04, 2012 at 06:31:06PM +0200, Florian Smeets wrote:
> On 04.09.12 18:14, Alex Kozlov wrote:
> > On Tue, Sep 04, 2012 at 12:32:44PM +0100, Chris Rees wrote:
> >> On 4 Sep 2012 05:15, "Alex Kozlov" <ak at freebsd.org> wrote:
> >>> On Mon, Sep 03, 2012 at 08:37:20PM -0700, Doug Barton wrote:
> >>>> On 09/03/2012 16:51, Florian Smeets wrote:
> >>>>> So I reread the complete thread (r301039 and followups) again, the
> >>>>> only problem there seems to be is pchecker complaining,
> >>>> It's a problem for anything that uses 'while read ...' to process
> >>>> those files, including portmaster.
> >>> I think it breaks all parsers that assume that distinfo is a text file
> >>> (have EoL on every line including last).
> >>>
> >>>> If you have the list of files with this problem and aren't interested
> >>>> in fixing them, pass the list along and I'll do it.
> >>> I plan to fix all occurrences of this in ports tree once every few months,
> >>> but if you wish to fix it now, here is a list:
> >>> audio/pianobar/distinfo
> >>> devel/pecl-libevent/distinfo
> >>> www/xpi-firebug/distinfo
> >>> www/xpi-showip/distinfo
> >>> www/trac-accountmanager/distinfo
> >>> mail/meta1/distinfo
> >>> net/istgt/distinfo
> >>> sysutils/fusefs-rar2fs/distinfo
> >> Is this a diff/patch bug?
> >> Perhaps it could be fixed as well... or could its behaviour be thought of
> >> as correct?
> > I tried many times but could not find the culprit. But I also suspect patch.
> No i don't think so, it is already in the diff you download from GNATS.
> I suspect it is a bug in getpatch, or one of the functions it calls, in
> query-pr.cgi. If you look at the patch in the PR with edit-pr (vi or
> vim) and use :set list you will see that the newline is there. This is
> why I emailed Shaun, so let's wait what he has to say.
It can be one of the sources, but in general it's not so simple.
For example update for sysutils/fusefs-rar2fs (ports/170258) was submitted
as shar file, so I created patch myself. I'm pretty sure, that shar created
files with proper EoL at last line and that patch also had proper EoL
(in fact, I just checked this).
There can be many sources of malformed patches, so I think that the best course
of action would be to add check for this to our patch. 
 
> BTW this does not only affect distinfo, it's always the last file of the
> patch, so there are a lot of patch-* files which have a missing newline
> at the end.
Here is the list (I marked binary files): http://people.freebsd.org/~ak/noeollist.txt


-- 
Alex