[PATCH xkbcomp 2/2] Fix incorrect processing of char(0xff) in comments

[Matthieu Herrb]

On Sun, Jan 16, 2011 at 12:27:30PM +0100, Dirk Wallenstein wrote:
> On Sun, Jan 16, 2011 at 01:26:10AM +0500, Alexandr Shadchin wrote:
> > If type readBuf is char, then scanchar() return -1 for char(0xFF) and EOF.
> > If type readBuf is unsigned char, then scanchar() return 0xFF for char(0xFF)
> > and -1 for EOF.
> 
> NAK: I assume you want to correct the case when returning EOF stored
> inside of readBuf?  This would now return EOF as 0x00ff but EOF is a
> negative value.  All parts that access readBuf work with signed
> values.

There are valid input files which include the 0xff character in
comments (the symbols/pc/fr file of the old xkb-data package).

whith the current signed implementation there's no way to make the
difference with EOF. 

I agree with Julien that reverting
9887842e33174a6e98675b912e419c711dd20c0f (and
41b0857c8c1179b87a26888588cbfff28f8bb0d6 and
61b3d4b024d8146cb0e7659aa958045ceb72c482
done afterwards to fix the initial commit) is probably better.

-- 
Matthieu Herrb