[systemd-devel] [PATCH] localed: validate set-x11-keymap input

[Lennart Poettering]

On Tue, 11.11.14 13:29, Jan Synacek (jsynacek at redhat.com) wrote:

> >> Hmm, reading the file over and over and over again sounds less than
> >> ideal. Maybe we should intrdouce a struct here and then make
> >> xkb_get_keymaps() return an array of structs really?
> >
> > That sounds ok, I'll see what I can do. I wanted to preserve as much of
> > the original code as I could, but maybe it wasn't the right decision.
> 
> After thinking about it, I can put the layout and its variants into a
> structure, but... When parsing "/usr/share/X11/xkb/rules/base.lst", is
> it safe to depend on the ordering of all the components in the file? I
> mean, is it safe to expect layouts being order before variants in that
> file? Or can it be vice versa? The code would have to be more general if
> one cannot depend on that order, and I think it would be quite ugly. The
> strv implementation doesn't have that problem.

Writing a parser that doesn't mind about the order doesn't appear too
complex to me... Would need a simply state machine that changes state
each time one of the "!" headers is encountered. And then, maintain a
"Set" for each type of field, and keep adding entries to it, ignoring
duplicates. And if you iterate through the variants and find a layout
that you haven't seen before just add the layout to the set, too.

Lennart

-- 
Lennart Poettering, Red Hat