[PATCH 2/2] Use GLib allocation in the audio subsystem.

[Boris Dusek]

Hello Andrei,

On Oct 4, 2010, at 12:05 PM, Andrei Kholodnyi wrote:

>> I will look at the things where we use GLib and try implementing
>> the small(er) things.
> 
> I have a question.
> Do you plan to remove GLib, if yes, then why.
> The reason I'm asking it was planned to switch to gmainloop,
> now, if it is still the case, why do you need to replace glib calls.

thanks for your question. It's a good one and I have been thinking about it
myself in the last 2 days.

My motivation to remove glib is that it is a big dependency on non-FreeDesktop
platforms. E.g. on Mac OS X, it's 1.0 MB (with single architecture) that I
would have to compile and redistribute with Speech Dispatcher. Whole Speech
Dispatcher itself is like 0.5 MB at most.

I thought of different options: reimplementing most GLib stuff we use in
libcommon, reimplementing all GLib stuff we use in libcommon, and using C++
standard library.  But in a situation which you pointed out (where we are
looking forward to extend usage of GLib for more important things like
mainloop, DBus etc., so GLib is here to stay), all these options start to seem
as redundant work. Also what this would lead to would be touching working code
and requiring people to change their habits (e.g. using spd_free instead of
g_free etc.).

I also considered as an important reason that FreeDesktop platform is an
absolute priority for Speech Dispatcher, and non-FreeDesktop platforms are
nice-to-have, but not priority, platforms.

So to sum up, I think that while having some negatives, when I consider all
things, then to go with glib is a reasonable way to go with the current
codebase of Speech Dispatcher.

Best regards,
Boris