Re: random commits to ORBit stable

[Sebastian Wilhelmi <wilhelmi ira uka de>]

Hi Michael,

> 	Trying to track down a bug in ORBit-stable on Power PC was ( and is )
> made extremely, tediously, soul wrackingly difficult by the lack of a
> suitable ChangeLog entry.
> 
> 	Furthermore I don't recall seeing the broken patch discussed here - it
> looks like it just 'magically' appeared in CVS. This is _really_
> under-appreciated in a stable branch.
>
> 	Can someone explain _why_ we need this optimization, where it was
> discussed, and why it was committed ?

The relevant mail is at

http://mail.gnome.org/archives/orbit-list/2001-March/msg00053.html

Afterwards Morten fixed the bug, I introduced with that patch (cough,
cough). That's the above cited patch. 

The whole point of this patch is to not decide at IDL-compile-time,
which memory has to be freed in the skels, but at run-time, just because
at compile-time you can't decide it. It's now decided by looking at the
4 bytes just before the string to free. It is either FREE_MARKER, in
which case the string gets freed, or it is the length of the current
string (i.e. some other number), in which case the string lies in the
GIOPRecvBuffer, and must not be freed.

> 	Also - it would be greatly appreciated if those ChangeLog haters would
> reform themselves and write proper ChangeLog entries.	

I'm indeed a ChangeLog lover, but adding an entry in a project without a
ChangeLog is quite pointless, I think.

Bye,
Sebastian
-- 
Sebastian Wilhelmi
mailto:wilhelmi@ira.uka.de
http://goethe.ira.uka.de/~wilhelmi