Re: [PATCH] Making XCB less of a memory hog

On Fri, 11 May 2007, Tilman Sauerbeck wrote:

> Barton C Massey [2007-05-11 13:28]:
> > In message <20070511191529.GA7797-pXKshE1ckPq3FNxj4QCNOA@xxxxxxxxxxxxxxxx> 
> > you wrote:
> > > Barton C Massey [2007-05-11 11:55]:
> > > > In message 
> > > > <20070511151628.GA1818-pXKshE1ckPq3FNxj4QCNOA@xxxxxxxxxxxxxxxx> you 
> > > > wrote:
> > > > > OTOH, this unfortunately breaks the API.
> > > >
> > > > My reading-XML-at-the-terminal ability is limited.  What
> > > > does the proposed API look like?  Or did you mean an ABI
> > > > break, about which we probably care not at all?
> > > [...] 
> > > So that ext pointer is moved from the struct to the function call.
> > > This would break the API and the ABI.
> > > 
> > > That's for the patch I submitted, but I also outlined a way to avoid the
> > > break, by adding xcb_send_request2(), that takes the new arguments.
> > > 
> > > Let me know if I need to explain better.
> > 
> > No, that gets it: thanks!  So the only API/ABI that breaks
> > is that between XCB extensions (and Xlib, as you point out)
> > and the XCB core.
> > 
> > In general I'm loathe to work around performance bugs in the
> > current implementation of shared libraries by distorting
> > XCB.  However, this seems like a relatively minor change;
> 
> Yeah, my timing really sucks. I wish I had noticed this behaviour when
> you were in the RC phase :(
> 
> Regards,
> Tilman
> 
> 

I say let's nip this in the bud.  There aren't that many consumers of XCB 
out there yet.