What's wrong with xkbcomp?

[Dan Nicholson]

On Tue, Sep 23, 2008 at 4:18 PM, Daniel Stone <daniel at fooishbar.org> wrote:
> On Tue, Sep 23, 2008 at 09:35:12PM +0100, Simos Xenitellis wrote:
>> An issue that is often discussed is the xkbcomp utility, that parses
>> the xkeyboard-config configuration files
>> and produces a binary XKM file that X.Org can read and take the
>> keyboard layout settings from there.
>>
>> Apparently, xkbcomp has performance issues when running on small devices.
>> On modern computers, xkbcomp is very fast. In the process of
>> diagnosing where the problem is,
>> I would like to ask for a sample of command-line options
>> that are used when you try to run xkbcomp on a small device.
>>
>> So, what parameters do you give to xkbcomp when you invoke it on these
>> small devices?
>
> setxkbmap -print | xkbcomp -xkb - foo.txt 2>/dev/null still takes 0.1sec
> on my laptop, with a completely warm cache.  With a cold cache, it takes
> around 1.3sec.  Given that we currently run this twice on startup, that's
> 1.4sec just compiling keymaps.  Ugh.
>
> Plus, why do we fork an external program, feed it a text description of
> five strings which is then fully lexed, have that generate an
> XkbDescRec, serialise the XkbDescRec to a binary format (losing
> information in the process) somewhere in the filesystem, read that and
> then deserialise to ... an XkbDescRec?

I thought someone (you or keithp?) was going to just put xkbcomp into
the server. Did anything ever happen with that? Just curious.

--
Dan