[systemd-devel] [PATCH] localed: forward xkbcommon errors

[Jan Synacek]

Lennart Poettering <lennart at poettering.net> writes:
> On Tue, 25.11.14 10:01, David Herrmann (dh.herrmann at gmail.com) wrote:
>
>> I explicitly ignore errors from verify_xkb_rmlvo() and proceed.
>> libxkbcommon is still not 100% compatible to libxkb (and doesn't
>> intend to be that, I guess). As we write X11 configs here, I just
>> continue with a warning.
>> 
>> If you call bus_error_setfv(), then sd-bus will return a method-error
>> to the caller. However, you also send a method-return in
>> method_set_x11_keyboard(). You thus end up with 2 calls.
>> 
>> I'm not sure how to solve this. Furthermore, libxkbcommon can print a
>> lot of informational warnings, and we shouldn't use just the last one.
>> One idea would be to copy the same logic into localectl. You could
>> also specify XKB_LOG_LEVEL and XKB_LOG_VERBOSITY as environment to get
>> more information there.
>> Yes, this would mean compiling the keymap twice, but it's for the sake
>> of debugging output so I think it's fine.
>
> What precisely are the error messages libxkbcommon prints? Are they
> really material to pass to the client? I mean, are they really about
> some invalid data the client passed or just about something broken in
> the way the system is set up? If it's the latter than I figure we
> should return just a simple error to the client, and pass the full
> logs to the logging stream...
>
> Lennart

Thinking about it again, I don't think that it's a good idea to pass
those errors to the client...

A few examples:

# localectl set-x11-keymap QQ

Dec 01 12:02:06 fedora-rawhide-systemd-virt systemd-localed[877]: libxkbcommon: Couldn't find file "symbols/QQ" in include paths
Dec 01 12:02:06 fedora-rawhide-systemd-virt systemd-localed[877]: libxkbcommon: 1 include paths searched:
Dec 01 12:02:06 fedora-rawhide-systemd-virt systemd-localed[877]: libxkbcommon:         /usr/share/X11/xkb
Dec 01 12:02:06 fedora-rawhide-systemd-virt systemd-localed[877]: libxkbcommon: Abandoning symbols file "(unnamed)"
Dec 01 12:02:06 fedora-rawhide-systemd-virt systemd-localed[877]: libxkbcommon: Failed to compile xkb_symbols
Dec 01 12:02:06 fedora-rawhide-systemd-virt systemd-localed[877]: libxkbcommon: Failed to compile keymap
Dec 01 12:02:06 fedora-rawhide-systemd-virt systemd-localed[877]: Cannot compile XKB keymap for new x11 keyboard layout ('' / 'QQ' / '' / ''): Invalid argument

# localectl set-x11-keymap us pc105 QQ

Dec 01 12:03:03 fedora-rawhide-systemd-virt systemd-localed[896]: libxkbcommon: Couldn't process include statement for 'us(QQ)'
Dec 01 12:03:03 fedora-rawhide-systemd-virt systemd-localed[896]: libxkbcommon: Abandoning symbols file "(unnamed)"
Dec 01 12:03:03 fedora-rawhide-systemd-virt systemd-localed[896]: libxkbcommon: Failed to compile xkb_symbols
Dec 01 12:03:03 fedora-rawhide-systemd-virt systemd-localed[896]: libxkbcommon: Failed to compile keymap
Dec 01 12:03:03 fedora-rawhide-systemd-virt systemd-localed[896]: Cannot compile XKB keymap for new x11 keyboard layout ('pc105' / 'us' / 'QQ' / ''): Invalid argument

Those errors are just stupid. Maybe they would make sense if there was
also libxkbcommon code in front of me... But from the user's
perspective, they say nothing useful. I added the prefix so the errors
at least have some context.

My original patch would say something like "No such layout 'QQ'" in the
first case and "No variant 'QQ' for layout 'us'" in the second.

So, is it worth logging this information? Any other ideas?

Cheers,
-- 
Jan Synacek
Software Engineer, Red Hat
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 818 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/systemd-devel/attachments/20141201/37150dcb/attachment.sig>