bug#20193: 25.0.50; declarative type specification for D-Bus args

[Daiki Ueno]

Michael Albinus <michael.albinus@gmx.de> writes:

>> I was actually not sure about the standard usage of subfeatures, and
>> copied the logic from process.c, where they are defined as a plist, so
>> they can be tested as:
>>
>>   (featurep 'make-network-process '(:server t))
>
> featurep has the restriction, that you can test only one subfeature per
> call. That's why they have created just one "subfeature", being a list.

Thanks for the explanation.

>> I am working on this, but it is turning to be non-trivial.  So, I have
>> pushed it to a separate branch scratch/dbusbind-type-tests, branched off
>> from scratch/dbusbind-type.
>
> Fortunately, I had some time to look on this today. I've committed some
> changes to dbusbind.c, all of them rather cosmetical.

Nice fixes, thanks!

> And I'm asking myself, whether we shall rename `dbus-message-internal'
> and `dbus-message-internal-to-lisp' to `dbus--message-*', in order to
> emphasize their internal nature.

That sounds like a good idea.

> I have added two tests to `dbus-test04-create-message-parameters', both
> fail. The first one must pass; this feature works in the master
> branch.

This was caused by a double registration of `:signature' symbol.  Fixed as:
http://git.savannah.gnu.org/cgit/emacs.git/commit/?h=scratch/dbusbind-type-tests&id=777848833cc9ff40411b78ad107e755172a881b8

> For the second new test I'm not sure whether this is possible (the
> documentation doesn't speak about), but it looks natural to me.

Yes, this was actually a bug because of missing checks on the number of
required arguments after `:type'.  Fixed as:
http://git.savannah.gnu.org/cgit/emacs.git/commit/?h=scratch/dbusbind-type-tests&id=def5829c0769b142b3cc0d69a9ad58935a9f237f

Regards,
-- 
Daiki Ueno