If it helps, the relevant portions of xorg.conf are as follows:

Section "ServerLayout"
    Identifier     "Layout0"
    Screen      0  "Screen0"
    InputDevice    "Keyboard0" "CoreKeyboard"
    InputDevice    "Mouse0" "CorePointer"
EndSection

Section "InputDevice"
    Identifier     "Mouse0"
    Driver         "synaptics"
    Option         "Protocol" "auto-dev"
##    Option         "MatchIsTouchpad" "on"
##    Option         "MatchDevicePath" "/dev/input/event*"
#    Option         "Device" "/dev/input/mice"
    Option         "Device" "/dev/psaux"
EndSection

please attach your full Xorg.log from 1.11 and 1.12. Looking at the one change that may affect this behaviour between 1.5.0 and 1.5.1, I suspect that 1.5.0 already ignored this stint in the configuration and went to the fallback mode.

The snippet you posted here indicates that the xorg.conf entry fails but the touchpad is still picked up through the hotplugging mechanisms, so it should just work.

Created attachment 58864 [details]
XOrg 1.12 with Synaptics driver 1.5.0 (working with bug)

It's somewhat working but it's buggy as the mouse pointer is often getting reset to the top-left corner.

Created attachment 58865 [details]
XOrg 1.12 with Synaptics driver 1.5.1 (server fails to start)

It makes the X Server completely unusable, as it fails to start.

Peter, please forget X Server 1.11, as it is obsolete now and I do not have enough time to test on many versions. I have attached the complete logs for Synaptics 1.5.0 (working with bug) and 1.5.1 (fails to start) respectively.

It would be nice if you could have a look to see what's going on... Thanks.

Same problem here.

1.5.1 causes the X server to hang during start.

1.5.99.901 works fine.

1.5.0 + 7c0361d4ec6b1f1325cb6551d0ee2e7f5cfae15b + c861d4568c2c6cc56db8e683a8d7035bef711262 + a15af628154ac37f24896a30614f3be6ecdb65c4 + 9f9b55ab55ed5251c1607c59d8817231d076d82c works fine too.

So the problem seems to be caused by either 3356dcb8f20cbcff4c08ba6a856a6c8d41f659d7 or df427ef5cc45c33a4159b7852cd40da6f7cc4b94.

Will test more tomorrow.

*** Bug 47734 has been marked as a duplicate of this bug. ***

Bug 47654 showed what was wrong - missing function. I cherry-picked the patch to the synaptics-1.5-branch, please test 75550077e66cb8bdf15d6e1cf5b28f884f508fc5 so I can avoid releasing another broken version. Thanks

You probably mean #47734 but anyway.
I can confirm that the problem with 1.5.0 from https://bugs.freedesktop.org/show_bug.cgi?id=47654#c3 is not going away with 1.5.1 (+latest patch) and xorg-server-1.12.
I currently use HEAD with xorg-server-1.12 respectively 1.5.1 with xorg-server-1.11.4 with no problems.

(In reply to comment #9)
> You probably mean #47734 but anyway.
> I can confirm that the problem with 1.5.0 from
> https://bugs.freedesktop.org/show_bug.cgi?id=47654#c3 is not going away with
> 1.5.1 (+latest patch) and xorg-server-1.12.
> I currently use HEAD with xorg-server-1.12 respectively 1.5.1 with
> xorg-server-1.11.4 with no problems.

sorry, I'm having real troubles parsing this. Do I get this right:
1.12 + 1.5-branch is still broken (cursor jump)
git master + server 1.12 is fine
1.5.1 with 1.11.4 is fine

If there is a simple way of building latest git or 1.5.99.x releases without having to install mtdev, then I might be able to carry out a quick test, to check that it does not crash the server upon starting up and to see if any of them solves the pointer jump bug...

Unless mtdev is really always needed, then it should probably be made optional through a build-time configuration option.

(In reply to comment #10)
> (In reply to comment #9)
> > You probably mean #47734 but anyway.
> > I can confirm that the problem with 1.5.0 from
> > https://bugs.freedesktop.org/show_bug.cgi?id=47654#c3 is not going away with
> > 1.5.1 (+latest patch) and xorg-server-1.12.
> > I currently use HEAD with xorg-server-1.12 respectively 1.5.1 with
> > xorg-server-1.11.4 with no problems.

Sorry for my bad english but you got it right.

Also the cursor moves (by default) slower maybe that's somehow related.

Going back to version 1.4.1 does not solve the bug of the pointer autonomously and randomly jumping back to the top-left corner during movement.

The pointer movement problem is unrelated to the hang. I think 7c0361d4ec6b1f1325cb6551d0ee2e7f5cfae15b fixed this in master. A new bug report would be best.

> I think 7c0361d4ec6b1f1325cb6551d0ee2e7f5cfae15b fixed this in master.
Yeah, that's the fix for it, thanks!

> A new bug report would be best.
I'm no friend of bureaucracy and as all problems are solved (at least for me) it should be cherry-picked and it's time for 1.5.2 (imo).

It is usually not a good idea to conflate several issues in one bug.
I reported bug 47654 now.

Sorry, I meant bug 47770.

Yes, I confirm, the patch from commit id 7c0361d4ec6b1f1325cb6551d0ee2e7f5cfae15b fixes the pointer jump to the top-left corner:

http://cgit.freedesktop.org/xorg/driver/xf86-input-synaptics/patch/?id=7c0361d4ec6b1f1325cb6551d0ee2e7f5cfae15b

However please note that compared to the previous XINPUT ABI, the sensitivity and acceleration seem very high/fast, so perhaps you might want to add some sort of attenuator ?

And yes, the pointer jump problem should be reported in a separate issue and perhaps a new release be made. It all started from another issue (build error in 1.5.0, see https://bugs.freedesktop.org/show_bug.cgi?id=47498), that's why the report looks now a bit messy if you had not followed up the whole story...

A separate bug has been filed for the pointer jump problem due to upgraded XINPUT ABI:

https://bugs.freedesktop.org/show_bug.cgi?id=47772

Therefore, this report is now exclusively devoted to the server startup crash with caused by xf86-input-synaptics-1.5.1 (+ XServer 1.12), as suggested by Chi-Thanh Christopher Nguyen.

(In reply to comment #11)
> Unless mtdev is really always needed, then it should probably be made optional
> through a build-time configuration option.

mtdev is needed for some multitouch devices, notably those that don't have ABS_MT_SLOT (protocol A devices). mtdev transparently wraps both protocol A and protocol B so we don't have to care about it. adding a bunch of ifdefs is painful to maintain so mtdev will likely stay around.

Fixed in version 1.5.2, along with https://bugs.freedesktop.org/show_bug.cgi?id=47770. Thanks.