Proposed libX11 ABI break
Peter Hutterer
peter.hutterer at who-t.net
Sat Jun 27 04:06:34 PDT 2009
On Fri, Jun 26, 2009 at 03:46:26PM -0400, Eamon Walsh wrote:
> Peter Hutterer wrote:
> > This week, I ran into a wall trying to get XI2 events sorted out. Resolving
> > this may require a libX11 ABI break. This is NOT XI2 specific, it just
> > happens to be the first extension to run into this issue.
> > This is a rather lengthy email, please take the time and read it.
> >
> > == Problem ==
> > XNextEvent(Display*, XEvent*) requires the event to be 96 (32-bit) or 192
> > (64-bit) bytes. This is a side-effect of the definition of the XEvent
> > structure itself.
> >
> > XGE's long events may require structs that exceed this space. For XI2, the
> > solution I have used so far are pointers inside the respective XEvent struct
> > that point to other allocated memory. The client is required to call
> > XIFreeEventData() to ensure this extra memory is freed accordingly.
> >
> >
>
> Why don't we just not support returning XGE events from those old
> functions ?
This was the alternative towards the end of the previous email. To quote:
>> The only other solution I could come up with so far is to add XGENextEvent()
>> and friends as substitutes for XNextEvent & co. In this approach, XNextEvent
>> _never_ returns generic events, leaving existing clients ABI-safe.
>> XGENextEvent requires an argument of the cookie+data type.
>>
>> This approach obviously has a high cost to clients as they need to adjust
>> for this new API.
>>
>> [Note that some new API is unavoidable. For example, XCheckTypedEvent does
>> not work in light of generic events as they share the same type. Additional
>> XGECheckTypedEvent() APIs are required, but they are required either way.]
basically this means that anyone dealing with the new XI2 API (or future
extensions) needs to also switch the mainloop to the new XGENextEvent and
cohorts.
Cheers,
Peter
More information about the xorg-devel
mailing list