Re: [vile] Where are 'keywords' set for vilemode, keymode, etc.?

[Thomas Dickey]

On Fri, Feb 16, 2007 at 01:30:58PM +0000, Eur Ing Chris Green wrote:
> On Fri, Feb 16, 2007 at 07:20:24AM -0500, Thomas Dickey wrote:
> > On Fri, Feb 16, 2007 at 12:04:40PM +0000, Eur Ing Chris Green wrote:
> > > I have the same vile.keywords file on all my systems, I don't have any
> > > 'local' vile.keywords files, there's only the one in /xxx/vile/share.
> > > Thus the highlighting *should* be the same on all systems, but it
> > > isn't for some reason.
> > 
> > Using an external filter (see, they have uses), you can add a -vvv option
> > and see the keyword- and color-information at the beginning of the output
> > from the filter.
> > 
> Can you explain a bit more fully please.

Stuff like this - useful for debugging filters:

flt_make_symtab(vile)
insert_keyword(Action, BC1, 1)

...     name "Action"   attr "BC1"
insert_keyword(Comment, C1, 1)

...     name "Comment"  attr "C1"
insert_keyword(Error, RC2, 1)

That makes sure you're really reading the colors you thought you were,
and that nothing's overriding the colors later in initialization.

> 
> I have experimented a bit more by writing a little bit of C which
> exercises all my syntax highlighting colors except bright magenta BC5.
> 
> All colors are correct in xvile on both Linux and Solaris (I don't
> have xvile on BSD).
> 
> All the non-bold colors are correct everywhere in vile.
> 
> On two Linux and one BSD system vile displays one color wrong:-
>     BC0 (bold black) is displayed as BC4 (bold blue)
> 
> On Solaris all the bold colors are wrong:-

hmm - Solaris raises an issue...

>     BC0 (bold black) is displayed as BC4 (bold blue)
>     BC1 (bold red) is displayed as C1 (red)
>     BC4 (bold blue) is displayed as C4 (blue)
> 
> All the systems *can* display all colors using the standard escape
> sequences.
> 
> It seems that somehow vile is getting the bold colors confused, and
> gets confused in a different way on Solaris.  All are vile 9.5o
> compiled with "--with-perl --with-ncurses", except one Linux system
> which is vile 9.5i (but acts exactly the same).

What does infocmp show on each system?  Look at the ncv value.

For instance, if you were relying on this bogus antique from Sun's
collection, that would be the cause of the problem:

#       Reconstructed via infocmp from file: /usr/share/lib/terminfo/x/xtermc
xtermc|xterm terminal emulator (color) @(#)xterm.ti 1.3,
        am, km, mir, msgr, xenl,
        btns#3, colors#8, cols#80, it#8, lines#24, ncv#7,
        pairs#64,

(It apparently describes a variant of xterm which I've never used, from
the early 90's).
 
> I'll try compiling with tcap instead of ncurses.

That could be different - ncurses pays more attention to ncv, and will
not try to mix attributes with color when ncv says the combination
will not work.

-- 
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net