Bug 1338585 - installing mate-desktop breaks gnome3 for touchpads
installing mate-desktop breaks gnome3 for touchpads
Status: CLOSED ERRATA
Product: Fedora
Classification: Fedora
Component: mate-settings-daemon (Show other bugs)
24
Unspecified Unspecified
unspecified Severity unspecified
: ---
: ---
Assigned To: Wolfgang Ulbrich
Fedora Extras Quality Assurance
: Reopened
: 1365732 (view as bug list)
Depends On:
Blocks:
  Show dependency treegraph
 
Reported: 2016-05-22 23:31 EDT by Adam Goode
Modified: 2017-02-28 11:22 EST (History)
13 users (show)

See Also:
Fixed In Version: mate-settings-daemon-1.16.0-2.fc25 mate-settings-daemon-1.16.0-2.fc24 mate-settings-daemon-1.16.0-2.el7
Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of:
Environment:
Last Closed: 2016-12-16 15:59:13 EST
Type: Bug
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---


Attachments (Terms of Use)

  None (edit)
Description Adam Goode 2016-05-22 23:31:23 EDT
Description of problem:
The mate-desktop package installs a symlink from /usr/share/X11/xorg.conf.d/99-synaptics-mate.conf to 50-synaptics.conf which overrides 90-libinput.conf. This breaks the gnome3 settings for touchpads.


Version-Release number of selected component (if applicable):
mate-desktop-1.14.1-1.fc24.x86_64


How reproducible:
Always.


Steps to Reproduce:
1. dnf install mate-desktop
2. Login to a gnome session

Actual results:
Touchpad is driven by synaptics driver.

Expected results:
Touchpad is driven by libinput driver.


Additional info:
Comment 1 Wolfgang Ulbrich 2016-05-23 01:53:23 EDT
Mate desktop does not support libinput driver and needs synaptic driver for a working touchpad.
Sorry, i can't remove this!
There is a upstream report about but mate desktop isn't ported for using libinput.
https://github.com/mate-desktop/mate-control-center/issues/133
Comment 2 leigh scott 2016-05-23 04:15:02 EDT
(In reply to Adam Goode from comment #0)
> Description of problem:
> The mate-desktop package installs a symlink from
> /usr/share/X11/xorg.conf.d/99-synaptics-mate.conf to 50-synaptics.conf which
> overrides 90-libinput.conf. This breaks the gnome3 settings for touchpads.
> 
> 
> Version-Release number of selected component (if applicable):
> mate-desktop-1.14.1-1.fc24.x86_64
> 
> 
> How reproducible:
> Always.
> 
> 
> Steps to Reproduce:
> 1. dnf install mate-desktop
> 2. Login to a gnome session
> 
> Actual results:
> Touchpad is driven by synaptics driver.
> 
> Expected results:
> Touchpad is driven by libinput driver.
> 
> 
> Additional info:

gnome will need to readd/add a synaptics fallback

xorg have bumped synaptic conf file priority as well

https://cgit.freedesktop.org/xorg/driver/xf86-input-synaptics/commit/?id=59e5db025307404fbfbc82f2fb3fe91d6a3005d7

https://cgit.freedesktop.org/xorg/driver/xf86-input-synaptics/commit/?id=59e5db025307404fbfbc82f2fb3fe91d6a3005d7
Comment 3 Peter Hutterer 2016-05-23 04:27:57 EDT
(In reply to leigh scott from comment #2)
> gnome will need to readd/add a synaptics fallback

I disagree. Mate needs to add libinput support. synaptics is in maintenance mode and will disappear over time, there is no point wasting effort on it. all the future effort on input device handling will go into libinput.

> xorg have bumped synaptic conf file priority as well
> 
> https://cgit.freedesktop.org/xorg/driver/xf86-input-synaptics/commit/
> ?id=59e5db025307404fbfbc82f2fb3fe91d6a3005d7

as the commit message tries to get across, this is actually so that removing synaptics completely from the distribution and users' workstations is easier in the long term. right now we require things like the symlink mentioned in comment #0. with this change, synaptics can be enabled/disabled by simply installing it.
Comment 4 Wolfgang Ulbrich 2016-05-23 04:37:55 EDT
(In reply to Peter Hutterer from comment #3)
> (In reply to leigh scott from comment #2)
> > gnome will need to readd/add a synaptics fallback
> 
> I disagree. Mate needs to add libinput support. synaptics is in maintenance
> mode and will disappear over time, there is no point wasting effort on it.
> all the future effort on input device handling will go into libinput.
> 
> > xorg have bumped synaptic conf file priority as well
> > 
> > https://cgit.freedesktop.org/xorg/driver/xf86-input-synaptics/commit/
> > ?id=59e5db025307404fbfbc82f2fb3fe91d6a3005d7
> 
> as the commit message tries to get across, this is actually so that removing
> synaptics completely from the distribution and users' workstations is easier
> in the long term. right now we require things like the symlink mentioned in
> comment #0. with this change, synaptics can be enabled/disabled by simply
> installing it.
Well, but that does not change the fact that it is impossible to use Mate /Cinnamon/Gnome on a same notebook system with f24, until Mate/Cinnamon support libinput.
Honestly, as Mate upstream is located at debian, linuxmint and ubuntu, i don't think that will happen before libinput driver is moved from debian-testing to debian-stable.
Comment 5 Peter Hutterer 2016-05-23 05:25:39 EDT
(In reply to Wolfgang Ulbrich from comment #4)
> Well, but that does not change the fact that it is impossible to use Mate
> /Cinnamon/Gnome on a same notebook system with f24, until Mate/Cinnamon
> support libinput.

yes, but that is a gnome/mate only issue/decision and has nothing to do with the commit you linked to. the visible effect once the server is up and running is exactly the same.
Comment 6 Adam Goode 2016-07-07 23:52:20 EDT
What can be done here?

I just hit this again where the old synaptics driver grabbed the device and confused me for a while.
Comment 7 leigh scott 2016-07-08 14:49:20 EDT
(In reply to Adam Goode from comment #6)
> What can be done here?
> 
> I just hit this again where the old synaptics driver grabbed the device and
> confused me for a while.

gnome will need to readd/add a synaptics fallback
Comment 8 Peter Hutterer 2016-07-10 19:43:33 EDT
(In reply to leigh scott from comment #7)
> (In reply to Adam Goode from comment #6)
> > What can be done here?
> gnome will need to readd/add a synaptics fallback

This isn't a Fedora issue, you'll need to file this bug upstream against GNOME. Because it isn't simply a matter of adding a synaptics fallback in g-s-d, you'll also require potential changes to the control center for options that libinput exposes differently than synaptics.

getting mate/cinnamon to support the libinput driver is the more realistic option especially given that this isn't just an alternative driver, it's the long-term  replacement and this work needs to be done at some point anyway.
Comment 9 Daniel Miranda 2016-07-10 20:54:26 EDT
The issue isn't the touchpad support itself, but the fact that the {mate,cinnamon}-desktop packages are required by applications (in my case, Nemo) while bringing a secondary dependency that is not directly related to those applications.

Isn't it possible to split the {mate,cinnamon}-desktop package, in a way that allows the applications to depend on one the parts, without pulling the Synaptics driver, while installing the complete desktop environments depends on it normally? For example, by having a {mate,cinnamon}-desktop-base package containing only the files strictly necessary for the applications to run, but not the {mate,cinnamon} session per-se?
Comment 10 leigh scott 2016-07-11 02:33:59 EDT
(In reply to Daniel Miranda from comment #9)
> The issue isn't the touchpad support itself, but the fact that the
> {mate,cinnamon}-desktop packages are required by applications (in my case,
> Nemo) while bringing a secondary dependency that is not directly related to
> those applications.

I have no interest in making nemo independent from cinnamon-desktop, nemo requires it to build.
Comment 11 leigh scott 2016-07-11 02:43:12 EDT
(In reply to Peter Hutterer from comment #8)

> getting mate/cinnamon to support the libinput driver is the more realistic
> option especially given that this isn't just an alternative driver, it's the
> long-term  replacement and this work needs to be done at some point anyway.

Cinnamon upstream have no interest in adding libinput support for at least two years as they are based on a LTS ubuntu release.
As for me adding support to it myself this is also unlikely as it's a real strain keeping up with the pointless gtk3 changes, this isn't likely to end till they stop the madness and release gtk4.
Comment 12 Daniel Miranda 2016-07-11 02:52:35 EDT
Whether Cinnamon does or does not support libinput is irrelevant.

cinnamon-desktop, from my understanding and the way it is used, is not a package that represents the whole Cinnamon *desktop environment*, but only the minimal structure for Cinnamon *applications* to run. Those applications in no way depend on the Synaptics driver.

If the Cinnamon settings daemon and/or control center do not support libinput, then *those packages* should conflict with the libinput packages to ensure they are not used, or much better, install a X11 config fragment to ensure that is the case - instead of relying on the assumption that the Synaptics driver will always have the highest priority.

At minimum the Synaptics driver dependency should be made weak (probably a Recommends) to allow manual removal of the package. This is still a suboptimal solutions as many users will install a Cinnamon application and have a system-wide change of behaviour in something completely unrelated.
Comment 13 Wolfgang Ulbrich 2016-07-11 03:46:41 EDT
(In reply to Daniel Miranda from comment #9)
> The issue isn't the touchpad support itself, but the fact that the
> {mate,cinnamon}-desktop packages are required by applications (in my case,
> Nemo) while bringing a secondary dependency that is not directly related to
> those applications.
> 
> Isn't it possible to split the {mate,cinnamon}-desktop package, in a way
> that allows the applications to depend on one the parts, without pulling the
> Synaptics driver, while installing the complete desktop environments depends
> on it normally? For example, by having a {mate,cinnamon}-desktop-base
> package containing only the files strictly necessary for the applications to
> run, but not the {mate,cinnamon} session per-se?
No chance to do this. Most Mate packages depends on libmate-desktop/libmate.
So an installation of ie. caja will always pull in mate-desktop.

And why do you use nemo in gnome?
I thought nautilus is the greatest software in the world :)

As Peter Hutterer said file out a report against gnome and ask for reverting the drop of synaptics support because you want to use nemo in gnome.
This can be easily done in a few hours if they care about users.

And Mate is is working on support of libinput, but we're a small team and at the moment this is a fedora problem only, as libinput isn't in debian stable.
So whole the Mate developers aren't affected.
Comment 14 Daniel Miranda 2016-07-11 04:06:27 EDT
> No chance to do this. Most Mate packages depends on libmate-desktop/libmate.
> So an installation of ie. caja will always pull in mate-desktop.

I don't see what this has to do with my point. The Synaptics dependency *does not* need to be in cinnamon-desktop. The fact that it worked before it was there, and that applications still work when I replaced it with a dummy package show that. It's simply broken packaging. If some other component of Cinnamon needs it then it should have that dependency, and it should not be forced into *every* Cinnamon application when they do not, in fact, need it.

> And why do you use nemo in gnome? I thought nautilus is the greatest software in the world :)

I would appreciate if you would refrain from making snarky comments and focus on the technical point at hand, as is the purpose of this bugtracker.

> As Peter Hutterer said file out a report against gnome and ask for reverting the drop of synaptics support because you want to use nemo in gnome.
> This can be easily done in a few hours if they care about users.

I don't see why I would file a bug in GNOME due to what is clearly a packaging error in Cinnamon in Fedora. I am *not* asking to use libinput with Cinnamon - I am asking for the packages to deal with it in a correct manner that does not affect other desktop environments. Which takes mere seconds and involve no changes to the code of either GNOME or Cinnamon. Only the packaging needs to be changed. There is absolutely no reason for Cinnamon to require libinput to be disabled when it is not in control of the desktop environment.

> And Mate is is working on support of libinput, but we're a small team and at the moment this is a fedora problem only, as libinput isn't in debian stable.
> So whole the Mate developers aren't affected.

Again, libinput support *does not matter at all*. What matters is *which* of the Cinnamon packages attempts to ensure Synaptics is used, and *how* it is done. *Both* are currently undeniably broken in Fedora. The base desktop package, shared by all applications, is not the one that should do it, and the correct way is *not* just to force the installation of xorg-x11-drv-synaptics. 

I'll try to be as clear as possible.

a) Every user that has any Cinnamon program installed, be it just one application or the whole environment, will have the cinnamon-desktop package installed.
b) Only users that have the whole Cinnamon environment installed and in use have to avoid using libinput.
c) Those users that have the whole Cinnamon environment installed and in use have the cinnamon-settings-daemon and/or cinnamon-control-center packages installed. These are the components that, in fact, depend on xorg-x11-drv-synaptics.

Therefore, the packages that should depend on xorg-x11-drv-synaptics are cinnamon-settings-daemon and/or cinnamon-control-center, and *not* cinnamon-desktop. That is the only action I am requesting.
Comment 15 Wolfgang Ulbrich 2016-07-11 04:35:02 EDT
(In reply to Daniel Miranda from comment #14)
> > No chance to do this. Most Mate packages depends on libmate-desktop/libmate.
> > So an installation of ie. caja will always pull in mate-desktop.
> 
> I don't see what this has to do with my point. The Synaptics dependency
> *does not* need to be in cinnamon-desktop. The fact that it worked before it
> was there, and that applications still work when I replaced it with a dummy
> package show that. It's simply broken packaging. If some other component of
> Cinnamon needs it then it should have that dependency, and it should not be
> forced into *every* Cinnamon application when they do not, in fact, need it.
> 
> > And why do you use nemo in gnome? I thought nautilus is the greatest software in the world :)
> 
> I would appreciate if you would refrain from making snarky comments and
> focus on the technical point at hand, as is the purpose of this bugtracker.
> 

Don't tell me what i say here!!!

> > As Peter Hutterer said file out a report against gnome and ask for reverting the drop of synaptics support because you want to use nemo in gnome.
> > This can be easily done in a few hours if they care about users.
> 
> I don't see why I would file a bug in GNOME due to what is clearly a
> packaging error in Cinnamon in Fedora. I am *not* asking to use libinput
> with Cinnamon - I am asking for the packages to deal with it in a correct
> manner that does not affect other desktop environments. Which takes mere
> seconds and involve no changes to the code of either GNOME or Cinnamon. Only
> the packaging needs to be changed. There is absolutely no reason for
> Cinnamon to require libinput to be disabled when it is not in control of the
> desktop environment.
> 
> > And Mate is is working on support of libinput, but we're a small team and at the moment this is a fedora problem only, as libinput isn't in debian stable.
> > So whole the Mate developers aren't affected.
> 
> Again, libinput support *does not matter at all*. What matters is *which* of
> the Cinnamon packages attempts to ensure Synaptics is used, and *how* it is
> done. *Both* are currently undeniably broken in Fedora. The base desktop
> package, shared by all applications, is not the one that should do it, and
> the correct way is *not* just to force the installation of
> xorg-x11-drv-synaptics. 

> 
> I'll try to be as clear as possible.
> 
> a) Every user that has any Cinnamon program installed, be it just one
> application or the whole environment, will have the cinnamon-desktop package
> installed.
> b) Only users that have the whole Cinnamon environment installed and in use
> have to avoid using libinput.
> c) Those users that have the whole Cinnamon environment installed and in use
> have the cinnamon-settings-daemon and/or cinnamon-control-center packages
> installed. These are the components that, in fact, depend on
> xorg-x11-drv-synaptics.
> 
> Therefore, the packages that should depend on xorg-x11-drv-synaptics are
> cinnamon-settings-daemon and/or cinnamon-control-center, and *not*
> cinnamon-desktop. That is the only action I am requesting.

This report is about mate-desktop and not cinnamon, see headline.
Please stop posting here problems with cinnamon or nemo.
Comment 16 Daniel Miranda 2016-07-11 04:41:22 EDT
> Don't tell me what i say here!!!

I don't think questioning my reasoning for using any particular software or mocking me for using GNOME has any place in this bugtracker. This is a space for technical discussion about the bugs.

> This report is about mate-desktop and not cinnamon, see headline.
> Please stop posting here problems with cinnamon or nemo.

The problem is exactly the same. I used {mate,cinnamon} as a placeholder earlier, and I thought it was clear enough that I'm talking about either. Replace every instance of Cinnamon with Mate in my previous comment and it still applies.

I would really like to understand what is the reasoning for *not* moving the dependency to the specific package that actually needs it.
Comment 17 Wolfgang Ulbrich 2016-07-11 05:35:07 EDT
Ok, i step back from this report as you don't stop spamming it with packages which aren't related to mate-desktop.
Comment 18 Daniel Miranda 2016-07-11 06:14:27 EDT
> Ok, i step back from this report as you don't stop spamming it with packages  
which aren't related to mate-desktop.

I don't understand your refusal to acknowledge the problem and discuss the solutions. Everything I said applies to mate-desktop, in exactly the same way as it does to cinnamon-desktop. You can literally replace every mention I made of Cinnamon to MATE and *nothing changes*.

Here are the exact same lines in the spec files of both packages:

http://pkgs.fedoraproject.org/cgit/rpms/mate-desktop.git/tree/mate-desktop.spec#n67
http://pkgs.fedoraproject.org/cgit/rpms/cinnamon-desktop.git/tree/cinnamon-desktop.spec#n34

I resent your claim that I am spamming: I argued my proposed fix in good faith hoping to find at least a partial solution, and you simply refused to consider it, and belittled me for something completely unrelated. I ask you to please read my comments, and consider moving the Synaptics dependency to a package other than {mate,cinnamon}-desktop.
Comment 19 Peter Hutterer 2016-07-12 01:47:32 EDT
what's the mate equivalent to gnome-control-center? Because that's the only package that should require synaptics (and tbh even that's a stretch but I can understand why one would argue for that).
Comment 20 leigh scott 2016-07-12 02:33:59 EDT
Why should we care if gnome breaks, this issue has been reassigned to gnome and they haven't responded (do they care about users?).
Comment 21 Daniel Miranda 2016-07-12 02:37:05 EDT
> Why should we care if gnome breaks

Because software in Fedora should not cause conflicts with any other unless absolutely necessary. And it is not necessary for the most trivial {mate,cinnamon}-desktop packages to depend on the Synaptics driver.

Again, this is a *packaging error* in Cinnamon, and nothing else. It is *not* a GNOME bug. Please analyse the situation rationally and undo the reassignment.
Comment 22 leigh scott 2016-07-12 02:47:28 EDT
(In reply to Daniel Miranda from comment #21)
> > Why should we care if gnome breaks
> 
> Because software in Fedora should not cause conflicts with any other unless
> absolutely necessary. And it is not necessary for the most trivial
> {mate,cinnamon}-desktop packages to depend on the Synaptics driver.
> 

Gnome caused the issue by removing their synaptics support too early.

> Again, this is a *packaging error* in Cinnamon, and nothing else. It is
> *not* a GNOME bug. Please analyse the situation rationally and undo the
> reassignment.

What is the point of reassigning the issue back to Mate or Cinnamon?, if it is reassigned back we will close the issue as "Wont Fix"!
Comment 23 Daniel Miranda 2016-07-12 02:48:56 EDT
> Gnome caused the issue by removing their synaptics support too early.

No. MATE and Cinnamon caused the issue by forcing a system-wide change of the touchpad driver in a package that has nothing to do with actually managing touchpads.

> What is the point of reassigning the issue back to Mate or Cinnamon?, if it is reassigned back we will close the issue as "Wont Fix"!

I will have no choice but to appeal to FESCo then.
Comment 24 leigh scott 2016-07-12 03:04:07 EDT
(In reply to Daniel Miranda from comment #23)
> > Gnome caused the issue by removing their synaptics support too early.
> 
> No. MATE and Cinnamon caused the issue by forcing a system-wide change of
> the touchpad driver in a package that has nothing to do with actually
> managing touchpads.
> 

So you will stop bitching if I move the synaptics dep to cinnamon main package as it has the touchpad settings?

> > What is the point of reassigning the issue back to Mate or Cinnamon?, if it is reassigned back we will close the issue as "Wont Fix"!
> 
> I will have no choice but to appeal to FESCo then.

That's your choice, it was also your choice of installing nemo which broke gnome for you.
Comment 25 Daniel Miranda 2016-07-12 03:41:39 EDT
> So you will stop bitching if I move the synaptics dep to cinnamon main package as it has the touchpad settings?

You should do it because it is the technically correct choice. I'm not bitching, I'm attempting to show that to you, but you and Wolfgang decided, without actually understanding my points, that you would not do it simply because I am a GNOME user, or because you believe GNOME should still support Synaptics.

Neither of those things has any bearing on what the *correct package to own the dependency* is. That's all I've been attempting to show all along.

> That's your choice, it was also your choice of installing nemo which broke gnome for you.

Installing an application that doesn't manage touchpads should not make it impossible to choose my touchpad driver. It's a completely reasonable expectation to have. In an ideal world it would be possible for the Synaptics override to apply only to MATE and Cinnamon, but I couldn't think of a way to do it. That being the case, the override should be restricted to the minimal set of packages possible, and that is *certainly not* "every single MATE or Cinnamon application".

---

I believe the best course of action would be to:

1) Move all the Synaptics-related packaging stuff to the packages that actually manages touchpad configuration. I believe they are {mate,cinnamon}-settings-daemon. Am I correct?
2) Make the driver override possible to disable. The two means that come first to my mind are:
a) Ship the override symlink in /etc/ instead of /usr/share, such that users can simply delete it
b) Ship the override in a separate package, that is pulled as a `Recommends` dependency by {mate,cinnamon}-settings-daemon, such that it is pulled by default, but can be uninstalled. Similar to the `dracut-config-*` packages, if an example is desired.
Comment 26 Bastien Nocera 2016-07-12 05:29:36 EDT
(In reply to leigh scott from comment #22)
> (In reply to Daniel Miranda from comment #21)
> > > Why should we care if gnome breaks
> > 
> > Because software in Fedora should not cause conflicts with any other unless
> > absolutely necessary. And it is not necessary for the most trivial
> > {mate,cinnamon}-desktop packages to depend on the Synaptics driver.
> > 
> 
> Gnome caused the issue by removing their synaptics support too early.

libinput was made the default for X.org 3 releases ago (in F22):
https://fedoraproject.org/wiki/Changes/LibinputForXorg

You will need to support the libinput Xorg drivers at some point.

In the meanwhile Daniel, you might want to try the "GNOME on Wayland" session, which will happily live with the synaptics driver being installed.
Comment 27 Daniel Miranda 2016-07-12 05:52:25 EDT
I actually want to use libinput *and* using MATE/Cinnamon apps, so I built a dummy package that provides xorg-x11-drv-synaptics just to satisfy the dependency, which is a really hacky solution.
Comment 28 Wolfgang Ulbrich 2016-07-12 06:35:48 EDT
Moving the force synaptics over libinput to another package does not fix the issue if an user or admin use several desktops on a server.
So, this isn't a solution and fix only your personal minimalistic point of view.

And again, this report is not about using nemo in gnome.
Header is about using different desktops and not a particular application.
So, feel free to file out tickets where ever you want.

Btw. there is also a report filed out from debian against gnome to revert the removal of synaptics support, until all desktops support libinput.
https://bugzilla.gnome.org/show_bug.cgi?id=768682

So reverting the change in gnome is the only solution which fixes the issue in all cases very shortly.
But gnome maintainer have re-asign it back, so i suspect there is no interest to fix your problem.


And Mate devs are working on support for libinput, but i can't beat them to work faster ;)
Comment 29 Daniel Miranda 2016-07-12 06:50:51 EDT
> Moving the force synaptics over libinput to another package does not fix the issue if an user or admin use several desktops on a server.

Yes, that's why I also suggested moving the override config file to /etc or a separate package. This way an administrator at least gets to choose which DE he prefers to work better while a common solution doesn't exist, or even handle input configuration manually through other means, such as `xinput` or Xorg config.

> And again, this report is not about using nemo in gnome.

It is about mate-desktop changing touchpad configuration. The original reporter has not mentioned whether they installed a single application or the whole environment. I just mentioned the way *I arrived* at the same issue.

I am still baffled by so much resistance to fixing a simple packaging mistake. The existence of other problems with libinput support does not justify it. There is no reason we need to commit another error on top of the problem that already exists. In fact, we should make all efforts to reduce the number of people affected as much as possible, not *extend it* to a bunch of users that are not actually using the MATE DE.
Comment 30 Wolfgang Ulbrich 2016-07-12 07:01:45 EDT
(In reply to Daniel Miranda from comment #29)
> > Moving the force synaptics over libinput to another package does not fix the issue if an user or admin use several desktops on a server.
> 
> Yes, that's why I also suggested moving the override config file to /etc or
> a separate package. This way an administrator at least gets to choose which
> DE he prefers to work better while a common solution doesn't exist, or even
> handle input configuration manually through other means, such as `xinput` or
> Xorg config.
> 
You're are complete wrong.
If i install  several desktops "complete" include mate or cinnmon, moving the 'prefer synaptics over libinput' will allways be installed.
I doesn't matter if i move this to another package.
A complete installation of MATE desktop will allways install it.
Comment 31 Daniel Miranda 2016-07-12 07:17:19 EDT
> You're are complete wrong.

I am waiting to be provided arguments for why reducing the users affected by the problem and offering admins the possibility of mitigation is wrong, when a permanent solution clearly isn't going to happen in the current Fedora release.

> If i install  several desktops "complete" include mate or cinnmon, moving the 'prefer synaptics over libinput' will allways be installed.

Yes, and if the symlink is moved to /etc it can be *removed* by the administrator. If it is moved to a different package that is an *optional* dependency of mate-settings-daemon that package can be *removed*. Right now it *cannot be removed* in any way. I'm proposing to solve what can be solved now.

> I doesn't matter if i move this to another package.

It matters because this situation will at least stop affecting *some people*, even though we can't solve it for *everyone*. If it is done *in addition* to making the override optional, there is a meaningful improvement.

I'm repeating myself again and again. I've said the same things three or four times. I'm done, I've been as clear as I can possibly be. I hope the maintainers or a proven packager do the right thing.
Comment 32 Wolfgang Ulbrich 2016-07-12 07:30:56 EDT
(In reply to Daniel Miranda from comment #31)
> Yes, and if the symlink is moved to /etc it can be *removed* by the
> administrator. If it is moved to a different package that is an *optional*
> dependency of mate-settings-daemon that package can be *removed*. Right now
> it *cannot be removed* in any way. I'm proposing to solve what can be solved
> now.

1. The link is in /etc
2. If an admin remove this link than touchpad support is broken for Mate desktop.

So, this is only a personal solution for you and not a general for using Mate and Gnome desktop on the same box.
Comment 33 Daniel Miranda 2016-07-12 07:36:23 EDT
The link is in /usr/share/X11/xorg.conf.d. If it wasn't the solution to go back to libinput would be to just delete it.

$ rpm -qlp mate-desktop-1.14.1-3.fc24.x86_64.rpm
/usr/bin/mate-about
/usr/bin/mate-color-select
/usr/bin/mpaste
/usr/share/X11/xorg.conf.d/99-synaptics-mate.conf
...

> So, this is only a personal solution for you and not a general for using Mate and Gnome desktop on the same box.

Only if the package grants me a special way to delete the symlink that doesn't apply to anyone else. Placing it /etc certainly will not do that.
Comment 34 Peter Hutterer 2016-09-01 21:27:01 EDT
(In reply to Wolfgang Ulbrich from comment #32)
> 1. The link is in /etc
> 2. If an admin remove this link than touchpad support is broken for Mate
> desktop.

well, yes. but that's not much different to the current solution where touchpad support is broken in gnome as soon as mate/cinnamon are *installed*. Placing the link in etc would at least give the decision on which desktop is broken to the admin.

> So, this is only a personal solution for you and not a general for using
> Mate and Gnome desktop on the same box.

There is no general solution as long as the two desktops do not support the same X input driver. There is no facility to change X drivers on the fly, so for now the symlinking is the only solution. So what we're looking for is a solution that
can be changed easily by the administrator after install. Or by avoiding the install altogether.
Comment 35 Wolfgang Ulbrich 2016-09-02 00:26:43 EDT
(In reply to Peter Hutterer from comment #34)
> (In reply to Wolfgang Ulbrich from comment #32)
> > 1. The link is in /etc
> > 2. If an admin remove this link than touchpad support is broken for Mate
> > desktop.
> 
> well, yes. but that's not much different to the current solution where
> touchpad support is broken in gnome as soon as mate/cinnamon are
> *installed*. Placing the link in etc would at least give the decision on
> which desktop is broken to the admin.
> 
> > So, this is only a personal solution for you and not a general for using
> > Mate and Gnome desktop on the same box.
> 
> There is no general solution as long as the two desktops do not support the
> same X input driver. There is no facility to change X drivers on the fly, so
> for now the symlinking is the only solution. So what we're looking for is a
> solution that
> can be changed easily by the administrator after install. Or by avoiding the
> install altogether.

I had a hard time i real life an was in hospital.
Sorry, i can't change anythink yet, maybe after f25 release my personal situation is better.
Peter, if you are a proven packager please move the link from /usr/share to /etc/ .
Package name is mate-desktop.
Comment 37 Fedora Update System 2016-09-04 23:10:09 EDT
mate-desktop-1.14.1-5.fc24 has been submitted as an update to Fedora 24. https://bodhi.fedoraproject.org/updates/FEDORA-2016-286001d03b
Comment 38 Fedora Update System 2016-09-04 23:10:42 EDT
mate-desktop-1.15.1-4.fc25 has been submitted as an update to Fedora 25. https://bodhi.fedoraproject.org/updates/FEDORA-2016-050df6c07a
Comment 39 Peter Hutterer 2016-09-04 23:10:52 EDT
done, I've moved the symlink to /etc! hope you're getting better or your situation is improving soon.

rawhide: http://koji.fedoraproject.org/koji/taskinfo?taskID=15501527
f25: http://koji.fedoraproject.org/koji/taskinfo?taskID=15501637
f24: http://koji.fedoraproject.org/koji/taskinfo?taskID=15501655

updates submitted where needed and they'll spam this bug soon
Comment 40 Fedora Update System 2016-09-05 15:22:09 EDT
mate-desktop-1.15.1-4.fc25 has been pushed to the Fedora 25 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2016-050df6c07a
Comment 41 Peter Hutterer 2016-09-05 21:12:27 EDT
leigh: I'm gonna push the same change into cinnamon-desktop, ok?
Comment 42 Fedora Update System 2016-09-05 23:21:27 EDT
mate-desktop-1.14.1-5.fc24 has been pushed to the Fedora 24 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2016-286001d03b
Comment 43 leigh scott 2016-09-06 01:37:19 EDT
(In reply to Peter Hutterer from comment #41)
> leigh: I'm gonna push the same change into cinnamon-desktop, ok?

Yes that's fine.

Thank you.
Comment 44 Fedora Update System 2016-09-06 02:33:39 EDT
cinnamon-desktop-3.0.2-2.fc25 has been submitted as an update to Fedora 25. https://bodhi.fedoraproject.org/updates/FEDORA-2016-0d27217da0
Comment 45 Fedora Update System 2016-09-06 03:16:53 EDT
cinnamon-desktop-3.0.2-2.fc24 has been submitted as an update to Fedora 24. https://bodhi.fedoraproject.org/updates/FEDORA-2016-177605da98
Comment 46 Peter Hutterer 2016-09-06 05:17:17 EDT
Having just discussed this with Benjamin, there's a better solution (and probably exactly what Daniel said out in comment #9):

let's make a mate-synaptics-symlink.fc25 package. That package installs the symlink and Requires xorg-x11-drv-synaptics. Then we can make mate-settings-daemon and mate-control-center Requires that package and remove the synaptics requirement from mate-desktop.

So for anyone installing mate-control-center or mate-settings-daemon everything remains the same. But those only installing an app that requires mate-desktop (like Nemo, apparently) won't get the symlink and have their desktop work fine.

That seems like a low-cost solution that can be immediately implemented and get everyone a bit closer to happiness.

The same applies to cinnamon with different package names.
Comment 47 Wolfgang Ulbrich 2016-09-06 08:23:32 EDT
Did you discussed the direct and and indirect dependencies to libmatedesktop from mate-desktop too ?  ;)
Only a few mate packages are independent from mate-desktop-libs, but not m-s-d and m-c-c.
So adding a extra package doesn't make any different.
Also, mate-desktop-libs is the package where i did all configuration for Mate.
And this is good and should not change.
With the update ppl who use Mate applications in Gnome or admins can remove the symlink, this should be enough until Mate is ported to libinput.
All Applications like engrampa, mate-terminal, atril or pluma could be used inside gnome before this update without installing the whole mate stack, so all this was a very theoretical fix in the update.
If someone want to replace nautilus with caja i'm sorry to say that caja depends on libmate-desktop too.

And again, the report was about co-installing mate and gnome which can't be fixed until Mate use libinput or gnome revert the revert of the synaptics driver.
Comment 48 Fedora Update System 2016-09-06 23:21:40 EDT
cinnamon-desktop-3.0.2-2.fc25 has been pushed to the Fedora 25 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2016-0d27217da0
Comment 49 Peter Hutterer 2016-09-07 00:35:02 EDT
I think you're misunderstanding the plan. AFAICT, the only actual dependencies on the synaptics driver are mcc and msd. Everything else doesn't care which driver is loaded. So the goal is to make those two depend on the new package that installs the symlink and remove the symlink from mate-desktop. No other dependencies are changed, so whatever depends on libmatedesktop now will continue to do so.
Comment 50 Fedora Update System 2016-09-09 15:22:20 EDT
cinnamon-desktop-3.0.2-2.fc24 has been pushed to the Fedora 24 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2016-177605da98
Comment 51 Fedora Update System 2016-09-21 20:26:03 EDT
mate-desktop-1.14.1-5.fc24 has been pushed to the Fedora 24 stable repository. If problems still persist, please make note of it in this bug report.
Comment 52 Fedora Update System 2016-09-22 03:33:24 EDT
mate-desktop-1.15.1-4.fc25 has been pushed to the Fedora 25 stable repository. If problems still persist, please make note of it in this bug report.
Comment 53 Bastien Nocera 2016-09-22 18:53:40 EDT
*** Bug 1365732 has been marked as a duplicate of this bug. ***
Comment 54 Fedora Update System 2016-10-03 13:23:15 EDT
cinnamon-desktop-3.0.2-2.fc24 has been pushed to the Fedora 24 stable repository. If problems still persist, please make note of it in this bug report.
Comment 55 Fedora Update System 2016-10-09 14:51:37 EDT
cinnamon-desktop-3.0.2-2.fc25 has been pushed to the Fedora 25 stable repository. If problems still persist, please make note of it in this bug report.
Comment 56 Fedora Update System 2016-12-10 16:43:46 EST
mate-settings-daemon-1.16.0-2.el7 mate-desktop-1.16.1-2.el7 has been submitted as an update to Fedora EPEL 7. https://bodhi.fedoraproject.org/updates/FEDORA-EPEL-2016-94da20123a
Comment 57 Fedora Update System 2016-12-10 16:44:02 EST
mate-settings-daemon-1.16.0-2.fc25 mate-desktop-1.16.1-2.fc25 has been submitted as an update to Fedora 25. https://bodhi.fedoraproject.org/updates/FEDORA-2016-b9ee64865a
Comment 58 Fedora Update System 2016-12-10 16:44:10 EST
mate-settings-daemon-1.16.0-2.fc24 mate-desktop-1.16.1-2.fc24 has been submitted as an update to Fedora 24. https://bodhi.fedoraproject.org/updates/FEDORA-2016-c6d6cef760
Comment 59 Fedora Update System 2016-12-12 07:47:44 EST
mate-desktop-1.16.1-2.el7, mate-settings-daemon-1.16.0-2.el7 has been pushed to the Fedora EPEL 7 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-EPEL-2016-94da20123a
Comment 60 Fedora Update System 2016-12-12 08:22:56 EST
mate-desktop-1.16.1-2.fc24, mate-settings-daemon-1.16.0-2.fc24 has been pushed to the Fedora 24 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2016-c6d6cef760
Comment 61 Fedora Update System 2016-12-12 08:25:30 EST
mate-desktop-1.16.1-2.fc25, mate-settings-daemon-1.16.0-2.fc25 has been pushed to the Fedora 25 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2016-b9ee64865a
Comment 62 Fedora Update System 2016-12-16 15:59:13 EST
mate-desktop-1.16.1-2.fc25, mate-settings-daemon-1.16.0-2.fc25 has been pushed to the Fedora 25 stable repository. If problems still persist, please make note of it in this bug report.
Comment 63 Fedora Update System 2016-12-20 01:24:05 EST
mate-desktop-1.16.1-2.fc24, mate-settings-daemon-1.16.0-2.fc24 has been pushed to the Fedora 24 stable repository. If problems still persist, please make note of it in this bug report.
Comment 64 Sylvia Sánchez 2016-12-21 18:03:07 EST
Still persists. I had to uninstall everything related to Mate to get Gnome touchpad settings working again.
Comment 65 Peter Hutterer 2016-12-21 18:38:25 EST
Sylvia: you should be able to just remove mate-settings-daemon now and that should remove the symlinks. At least that's how it looks like from the spec file diff.
Comment 66 Wolfgang Ulbrich 2016-12-22 14:28:11 EST
[rave@f25-origin ~]$ sudo dnf remove mate-settings-daemon
Dependencies resolved.
================================================================================
 Package                          Arch     Version            Repository   Size
================================================================================
Removing:
 fortune-mod                      x86_64   1.99.1-22.fc24     @anaconda   2.8 M
 mate-applets                     x86_64   1.16.0-1.fc25      @anaconda    16 M
 mate-control-center              x86_64   1.16.0-1.fc25      @anaconda   7.3 M
 mate-control-center-filesystem   x86_64   1.16.0-1.fc25      @anaconda     0  
 mate-desktop                     x86_64   1.16.1-2.fc25      @System     183 k
 mate-dictionary                  x86_64   1.16.0-1.fc25      @anaconda    12 M
 mate-menu                        noarch   16.10.1-2.fc25     @System     1.3 M
 mate-panel                       x86_64   1.16.0-3.fc25      @System     9.1 M
 mate-panel-libs                  x86_64   1.16.0-3.fc25      @System      93 k
 mate-power-manager               x86_64   1.16.0-1.fc25      @anaconda   7.4 M
 mate-session-manager             x86_64   1.16.0-1.fc25      @anaconda   1.4 M
 mate-settings-daemon             x86_64   1.16.0-2.fc25      @System     2.6 M
 mate-system-monitor              x86_64   1.16.0-1.fc25      @anaconda   4.3 M
 recode                           x86_64   3.6-42.fc24        @anaconda   2.1 M
 upower                           x86_64   0.99.4-2.fc25      @anaconda   500 k
 xorg-x11-drv-synaptics           x86_64   1.9.0-1.fc25       @updates    132 k

Transaction Summary
================================================================================
Remove  16 Packages

Installed size: 68 M
Is this ok [y/N]: 

Well, i told you so that moving the link to m-s-d is pretty useless ;-)
M-s-d is one of main package for MATE and just married with some other packages.
You can remove the symlink by hand or wait for port to libinput by upstream.
I am pretty sure this will be happen before next stable debian release.
Comment 67 Fedora Update System 2016-12-27 12:17:32 EST
mate-desktop-1.16.1-2.el7, mate-settings-daemon-1.16.0-2.el7 has been pushed to the Fedora EPEL 7 stable repository. If problems still persist, please make note of it in this bug report.
Comment 68 Brian J. Murrell 2017-01-25 11:32:12 EST
So, granted this is a solution that only works with gdm (but of course can be ported to other DMs), but gdm knows what kind of session the user wants to start before an Xorg is started for it.

Can it not be hacked to set up /etc/X11/xorg.conf.d/ with the correct drivers for the desktop environment until this change has coalesced through the various desktops?

I suppose such a solution is ultimately a little racey (against two people logging into different Xorgs sessions at the same time) but in all practicality, it's not a race that should be hit very often if at all, ever.

Or can Xorg even be pointed at different /etc/X11/xorg.conf.d dirs when it's started and then such a dir can be created ephemerally when an Xorg is being started up and then thrown away.  Or Fedora can ship with per-desktop Xorg configuration dirs.
Comment 69 Wolfgang Ulbrich 2017-01-25 12:14:56 EST
MATE upstream added support for libinput.
Rawhide (f26) is already updated to use it.
http://pkgs.fedoraproject.org/cgit/rpms/mate-settings-daemon.git/commit/?id=1bb0c20e28491eb83ad7c09d11e436e64d821be9
http://pkgs.fedoraproject.org/cgit/rpms/mate-control-center.git/commit/?id=f3d348f34775b4fdf0cfc4c7d9814018bae62424

This is part of MATE-1.18 and i am optimistic to upgrade f25 after f26 is released.

Note You need to log in before you can comment on or make changes to this bug.