Messed-up encoding in doc-comments (in the popup-widget)

[Niko Sams]

On Wed, Jul 8, 2009 at 08:44, Andreas Pakulat<apaku at gmx.de> wrote:
> On 08.07.09 07:03:46, Niko Sams wrote:
>> On Wed, Jul 8, 2009 at 00:16, Andreas Pakulat<apaku at gmx.de> wrote:
>> > On 07.07.09 23:58:53, David Nolden wrote:
>> >> Am Dienstag 07 Juli 2009 14:53:59 schrieb Milian Wolff:
>> >> > Hey guys, I want to fix bug https://bugs.kde.org/show_bug.cgi?id=183182
>> >> > but am a bit lost. Can someone give me a bit of insight?
>> >> >
>> >> > What I found out is that it only affects doc-comments and that it is not
>> >> > due to formatComment().
>> >> >
>> >> > 1) How could I properly debug this stuff? kDebug() doesn't seem to work
>> >> > fine. Could I use GDB and somehow print me the text and see where it
>> >> > gets corrupted?
>> >> >
>> >> > 2) Could the popup-widget be the culprit? Where are its sources again?
>> >> Generally, the comments are supposed to be utf8 encoded within the duchan. And
>> >> kDebug() should work properly with them. If it doesn't, then that's probably
>> >> already part of the problem.
>> >>
>> >> There is one thing that comes into my mind. Look at
>> >> kdevelop/languages/cpp/preprocessjob.cpp: I think there's somewhere a @todo,
>> >> saying something like "convert the file to utf-8 if it isn't yet", and I think
>> >> that todo is there still.
>> >>
>> >> That would need a check like "If the local encoding is not utf-8, convert the
>> >> text before processing it".
>> >
>> > I'd just like to add that "use the local encoding" might not necessarily
>> > help. Most distro's use utf-8 as default encoding, but the user may have
>> > files from "back then" when he was using KOI-8, latin9 or whatever.
>> >
>> > As selecting the right encoding for each file is not going to work for the
>> > background parser, I'm wondering wether maybe we should try to use th
>> > KEncodingProber class from kdelibs (IIRC thats the name of the newer one,
>> > using the algorithms developed by mozilla to detect the encoding)?
>> I think a setting per project is needed. Katepart should also use that setting.
>> Autodetection can just guess. (for parsing this could be enough - but not
>> for editing)
>
> Well, I'm not sure a project-wide option helps, these people usually have a
> different encoding on every other file :)
Quanta 3 had that and that worked fine.

> Hmm, I think there's a way to specify an encoding via the
> .kate-command-file (not sure what the filename is, it takes the same
> kate-commands as the kate-modeline), so if we can somehow find and read
> that and just use that encoding the user at least has a way of specifying
> the encoding on a per-directory base.
That's perfect. We can even add a UI for that.

Niko