Revision 755d5487f554cf585f3b988088dc20fad65d1bf6

[Mark Doliner]

On Wed, 11 Jul 2007 16:27:15 +0400, Shkutkov Michael wrote
> Mark Doliner wrote:
> > On Tue, 10 Jul 2007 18:51:40 -0400 (EDT), mshkutkov wrote
> >   
> >> -----------------------------------------------------------------
> >> Revision: 755d5487f554cf585f3b988088dc20fad65d1bf6
> >> Ancestor: 1b455e6f5ef8d75f46d5faa291a761ec59269586
> >> Author: mshkutkov at soc.pidgin.im
> >> Date: 2007-07-10T21:19:49
> >> Branch: im.pidgin.soc.2007.remotelogging
> >>
> >> Modified files:
> >>         libpurple/connection.c
> >>
> >> ChangeLog:
> >>
> >> purple_connection_set_state was made nonblocking by using
> >>     
> > purple_log_write_nonblocking.
> >
> > I haven't been following the remotelogging branch, but this change seems weird
> > to me.  So will we need to change all code that calls purple_log_write() to
> > now call purple_log_write_nonblocking()?
> >
> > Couldn't we just change purple_log_write() to always be nonblocking? 
> > purple_log_write() could strdup the message, and then free it after it's been
> > written to the file.  So you could get rid of the 'cb' parameter and just
> > handle it internally in log.c.  That way the callers of purple_log_write()
> > wouldn't need to re-implement it every time.
> >
> > You could even optimize that a little bit so that purple_log_write() accepts a
> > true/false parameter that, when false it strdups the message, but when true it
> > takes ownership of the passed in message string and frees it when finished.
> >
> > -Mark
> >
> >   
> Having puprle_log_write function which call callback maybe useful,
> because in callback we know the result of writing message, was it 
> successful or not.
> So we can handle this information and for example the UI could 
> display a failure message.

I kinda feel like, if the logger isn't able to log something, then it is the
responsibility of the logger to show an error message.  And the caller of
purple_log_write() shouldn't have to deal with it.  Mostly I'm just wary about
adding an extra callback function everywhere to free the message.  What do
other people think?

> As C doesn't have function overloading, we could have several 
> writting log functions:
> 
> purple_log_write_with_cb
> purple_log_write_no_dup
> purple_log_write
> 
> If so, we can leave all code that uses old purple_log_write as it is.
> 
> P.S. _nonblocking suffix will be removed later.

Ah, I see.  That sounds good to me.

-Mark