[Bug 71093] finish making parameters fully typed

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=71093

--- Comment #24 from Simon McVittie <simon.mcvittie at collabora.co.uk> ---
(In reply to comment #13)
> [tp-glib master 1/2] tp_connection_manager_param_dup_variant_type: add
> shouldn't you annotate this (transfer full)?

Er, yes. I assume the telepathy-glib bits are OK with that change?

(In reply to comment #15)
> Humm wouldn't it be easier to return a struct/object then? Something like
> TpConnectionManagerParam so we can have method to check if the param is
> secret or not.

If we're going to allow for future expansion like this, I think I'd rather make
the vfunc return a list of names (strv of strings, probably) like
get_untyped_parameters() does, and iterate through that calling get_parameter()
for each one - that seems simpler than inventing objects. Would that be OK from
your point of view?

If we ever need metadata other than flags, we can have a HAS_METADATA flag and
an optional get_parameter_metadata(), or something.

(In reply to comment #4)
> The commit message doesn't say, but this patch also adds a get_features()
> function, because we need that to be able to say "this backend can store
> types". I wonder whether to rename it to get_flags() so we can use it for
> things that aren't necessarily "positive", like MIGRATED_FROM_0_x?

Having thought about this since last week, my conclusion is "yes, they should
be called Flags". I'll cut it out into its own patch, too.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.