Merge lp://staging/~bregma/geis/lp-741404 into lp://staging/geis

"This will restore buildability at the cost of device enumeration."

Stephen, could you elaborate a little more on the reason for device enumeration being "lost"?

And to clarify:

Device will continue to work if XI2.1 is available and support for it is compiled, yes?

Are the two ways to test for the hidden attribute equivalent? If so, perhaps it should be saved for oneiric?

> "This will restore buildability at the cost of device enumeration."
>
> Stephen, could you elaborate a little more on the reason for device
> enumeration being "lost"?

Absolutely.

Utouch-geis uses XI2.1 calls to the X server to enumerate the available multi-touch capable input devices. The functionality available through XI2.1 is an API-breaking change currently available only in Ubuntu 11.04 and later or those other distributions willing to port the Ubuntu-specific x.org patches.

The API breakage is in the x.org client library headers but also requires an ABI breakage in the entire x.org stack. Some distributions wish to take advantage of uTouch as it worked in Ubuntu 10.10 and the added functionality delivered by XI2.1 is not necessary for that.

This change makes the use of the API-breaking changes configurable at compile time for those downstream consumers who have been unable to build utouch-geis since version 2.0.4.

> And to clarify:
>
> Device will continue to work if XI2.1 is available and support for it is
> compiled, yes?

Functionality on Ubuntu will continue to work as before. Use of XI2.1 is enabled by default in development systems that support it unless steps are take to disable it with --disable-xi2.1 when running configure.

> Are the two ways to test for the hidden attribute equivalent? If so, perhaps
> it should be saved for oneiric?

I'm not sure I understand this question. Do you mean utouch-geis should talk to the kernel of the host on which it's running to determine the capabilities of local hardware as an alternative to querying the X server if that choice is not available at compile time? That's possible as an add-on, but the changes in this merge would still be required.

This change is not targeted at Ubuntu. It does not need to get in to natty. It is for other downstream consumers of utouch-geis.

I was referring to lines 33-38 in the diff. Either way, since this is general development, no problem.

Looks good to me too!

> I was referring to lines 33-38 in the diff. Either way, since this is general
> development, no problem.

Ah, I understand now.

AC_TRY_COMPILE is deprecated in autoconf in favour of AC_COMPILE_IFELSE. Since I was using the newer construct elsewhere I though I'd just go for consistency. The two compile tests are functionally equivalent in this case.