I'd like to know whether there are plans to make wayland default in F25. I'm asking because there was no change proposed for this¹, although Fedora Magazine is stating it will become default in F25².
I know that there have been some considerations on this before³ and some places where issues are collected⁴.
Still, the wiki page is missing some important stuff:
How do we handle the fact that even XWayland applications cannot be started as root? I'm ok with dropping the "feature" of starting GUI applications as root, but this might need some discussion and will need some communication. Please note that this would drop any support for at least a dozen applications such as GParted or grub-customizer. And, as Kamil Paral stated on the december thread:
I think we're going to get much worse publicity with this than primary selection. Inexperienced Linux users are used to run "sudo gedit", because terminal editors are unfriendly. Even I, with my years of experience, use "sudo meld" to merge configuration files, because I don't want to learn vimdiff, and graphical apps can simply offer a much superior experience. There was a discussion on devel list about this and everybody talked why running root gui apps is unsafe on X11, but nobody explained why it is unsafe on Wayland. Polkit and fine-grained permissions were proposed, which is a good idea, but it's not going to solve all use cases and it's not the current state of things in many apps. So unless there's a very good reason for deviating from current common practices, I'd put this onto "needs to be fixed" list. The more things we break without a really good reason, the more pushback we're going to get. Maybe this can be improved gradually over time?
This point should at least be noted on the wiki list. Can someone please add it?
Accessibility is pretty much unusable on wayland right now. The wiki page lets me assume that nobody really worked on this besides some minor bug fixings. Can we disable wayland for people relying on a11y? If not, shipping wayland by default probably is a bad idea.
Virtual machine and remote desktop software: Currently, it is impossible for remote desktop and virtual machine GUIs (both under XWayland and Wayland) to grab keys. This is a major regression for all users of these pieces of software.⁵ This point is completely missing on the wiki page. Can someone please add it?
Stability: There are more crashers in gnome-shell with wayland than with X11. That's probably ok, but still worth considering.
Drivers: Lastly, I don't know anything about support with non-Intel GPUs, because their drivers (especially proprietary ones) used to mess up pretty much last time I used them.
¹ https://fedoraproject.org/wiki/Releases/25/ChangeSet
² https://fedoramagazine.org/whats-new-fedora-24-workstation/
³ See e.g: * "Plan to organize the Wayland-by-default effort" in late 2015, https://lists.fedoraproject.org/archives/list/desktop@lists.fedoraproject.or... * "Wayland blockers for F24 default" in februar, https://lists.fedoraproject.org/archives/list/desktop@lists.fedoraproject.or...
⁴ See https://bugzilla.redhat.com/show_bug.cgi?id=WaylandRelated https://bugzilla.gnome.org/show_bug.cgi?id=WaylandRelated and https://bugzilla.gnome.org/show_bug.cgi?id=wayland https://fedoraproject.org/wiki/Wayland_features
⁵ https://bugs.freedesktop.org/show_bug.cgi?id=97333 and https://bugzilla.redhat.com/show_bug.cgi?id=1285770
On Mon, Aug 15, 2016 at 02:21:05PM -0000, Christian Stadelmann wrote:
Stability: There are more crashers in gnome-shell with wayland than with X11. That's probably ok, but still worth considering.
Except that a gnome-shell crash on Wayland makes all the applications to crash as well, AFAIK.
-- Sébastien
On Mon, Aug 15, 2016 at 02:58:21PM -0000, Christian Stadelmann wrote:
Except that a gnome-shell crash on Wayland makes all the applications to crash as well, AFAIK.
That's true, yes. I don't know whether it has to be this way though. In theory, applications could just reopen the socket and paint again, which might re-trigger the crash though.
It's a major regression in robustness…
-- Sébastien
On Mon, Aug 15, 2016 at 7:42 PM, Michael Catanzaro mcatanzaro@gnome.org wrote:
On Mon, 2016-08-15 at 18:39 +0200, Sébastien Wilmet wrote:
It's a major regression in robustness…
Yup. GNOME Shell crashes are not super common, but they still happen more than they should, so this is a serious point. Is there a bug for it?
Does this also affect the "Alt+F2 r" restart functionality?
On Mon, Aug 15, 2016 at 8:00 PM, Michael Catanzaro mcatanzaro@gnome.org wrote:
On Mon, 2016-08-15 at 19:46 +0100, Peter Robinson wrote:
Does this also affect the "Alt+F2 r" restart functionality?
Yeah, if you try it just says: "Restart is not available on Wayland"
Damn, such a useful feature, I use it quite a bit as a session that's up for a few days to a week gets sluggish
On Mon, Aug 15, 2016, 3:26 PM Peter Robinson pbrobinson@gmail.com wrote:
On Mon, Aug 15, 2016 at 8:00 PM, Michael Catanzaro mcatanzaro@gnome.org wrote:
On Mon, 2016-08-15 at 19:46 +0100, Peter Robinson wrote:
Does this also affect the "Alt+F2 r" restart functionality?
Yeah, if you try it just says: "Restart is not available on Wayland"
Damn, such a useful feature, I use it quite a bit as a session that's up for a few days to a week gets sluggish
Is there a bug for that? If this is due to the Shell (as it seems to be) that has to be fixed.
On Mon, 2016-08-15 at 19:46 +0100, Peter Robinson wrote:
Does this also affect the "Alt+F2 r" restart functionality?
Yeah, if you try it just says: "Restart is not available on Wayland"
Damn, such a useful feature, I use it quite a bit as a session that's up for a few days to a week gets sluggish
Is there a bug for that? If this is due to the Shell (as it seems to be) that has to be fixed.
No, primarily because I have no idea how you'd quantify that.
I've just filed a bug against gnome-shell: https://bugzilla.redhat.com/show_bug.cgi?id=1367666
Hi all,
----- Original Message -----
I've just filed a bug against gnome-shell: https://bugzilla.redhat.com/show_bug.cgi?id=1367666
Yes, I commented in this bug.
Basically, imho, gnome-shell on Wayland is not worse than the X server on X11.
Wayland and X11 do not have the same architecture (if they did, I guess we would not need need Wayland)
In X11, if the X server dies, the entire session goes along with it, whereas the window manager is just an X11 client, so it can be restarted at will. When restarting the window manager, the other X11 applications connected to the X server can continue to work as the X server remains the same.
In Wayland, if the Wayland compositor dies, the entire session goes along with it. In GNOME, gnome-shell is the window manager and the Wayland compositor, which means that losing gnome-shell means losing the session, just like losing the X server means losing the session in X11.
So indeed, you cannot restart gnome-shell in Wayland, it's pretty much by design, just like you cannot restart weston if you use weston as your Wayland compositor.
But is being able to restart gnome-shell such an important feature? Isn't that a debug (user hitting alt-f2 + "r", what's the real use case for that, in a normal, regular user session?) - What matters more, imho, is that gnome-shell is robust enough so that it doesn't crash.
Like any software, there can be bugs, just like the X server can have bugs as well and the X server do crash at times, sometimes it's due to a bug in the Xserver, sometimes it's a bug in one of the drivers, but the net result is basically the same, when that occurs in X11, the user loses his/her entire session. I don't see gnome-shell/Wayland being any worse than the Xserver on X11 in this regard. When bug occur, a bug is filed and the developers do their best to fix it.
However, unlike weston, gnome-shell requires Xwayland to run, so losing Xwayland means losing gnome-shell, which means losing the session. If we could get rid of this requirement on Xwayland in gnome-shell, gnome-shell/mutter could possibly survive a crash in Xwayland. That would be upstream bug https://bugzilla.gnome.org/show_bug.cgi?id=759538 but it's not trivial and requires quite a few changes in mutter as it still uses X11 heavily in multiple places.
Cheers, Olivier.
On Wed, Aug 17, 2016 at 10:10:42AM -0400, Olivier Fourdan wrote:
I've just filed a bug against gnome-shell: https://bugzilla.redhat.com/show_bug.cgi?id=1367666
Yes, I commented in this bug.
Basically, imho, gnome-shell on Wayland is not worse than the X server on X11.
Wayland and X11 do not have the same architecture (if they did, I guess we would not need need Wayland)
In X11, if the X server dies, the entire session goes along with it, whereas the window manager is just an X11 client, so it can be restarted at will. When restarting the window manager, the other X11 applications connected to the X server can continue to work as the X server remains the same.
In Wayland, if the Wayland compositor dies, the entire session goes along with it. In GNOME, gnome-shell is the window manager and the Wayland compositor, which means that losing gnome-shell means losing the session, just like losing the X server means losing the session in X11.
Are there statistics about the number of X server crashes in Fedora?
From my experience, the X server crashes far less often than gnome-shell.
-- Sébastien
Hi,
Are there statistics about the number of X server crashes in Fedora?
From my experience, the X server crashes far less often than gnome-shell.
Best source for such data is bugzilla, something along these lines:
Xwayland: https://bugzilla.redhat.com/buglist.cgi?bug_status=NEW&bug_status=ASSIGN...
Xorg: https://bugzilla.redhat.com/buglist.cgi?bug_status=NEW&bug_status=ASSIGN...
gnome-shell/mutter https://bugzilla.redhat.com/buglist.cgi?bug_status=NEW&bug_status=ASSIGN...
This can certainly be improved, for example, filter the gnome-shell/mutter bugs that apply to Wayland only.
HTH, Cheers, Olivier;
On Wed, Aug 17, 2016 at 10:10:42AM -0400, Olivier Fourdan wrote:
So indeed, you cannot restart gnome-shell in Wayland, it's pretty much by design, just like you cannot restart weston if you use weston as your Wayland compositor.
But is being able to restart gnome-shell such an important feature? Isn't that a debug (user hitting alt-f2 + "r", what's the real use case for that, in a normal, regular user session?) - What matters more, imho, is that gnome-shell is robust enough so that it doesn't crash.
This should probably be added to https://fedoraproject.org/wiki/Wayland_features even if it's not exactly a feature. (Both the ability to restart intentionally and the ability to recover from a shell crash with minimal user disruption.)
I can add it to the wiki if need be, although it's probably better if someone with domain expertise does.
Hi,
On Wed, Aug 17, 2016 at 10:10:42AM -0400, Olivier Fourdan wrote:
So indeed, you cannot restart gnome-shell in Wayland, it's pretty much by design, just like you cannot restart weston if you use weston as your Wayland compositor.
But is being able to restart gnome-shell such an important feature? Isn't that a debug (user hitting alt-f2 + "r", what's the real use case for that, in a normal, regular user session?) - What matters more, imho, is that gnome-shell is robust enough so that it doesn't crash.
This should probably be added to https://fedoraproject.org/wiki/Wayland_features even if it's not exactly a feature. (Both the ability to restart intentionally and the ability to recover from a shell crash with minimal user disruption.)
I can add it to the wiki if need be, although it's probably better if someone with domain expertise does.
Sure, it's now added to the Wayland feature page:
https://fedoraproject.org/wiki/Wayland_features#Restarting_gnome-shell
Cheers, Olivier
On Mon, Aug 15, 2016 at 3:21 PM, Christian Stadelmann genodeftest@fedoraproject.org wrote:
I'd like to know whether there are plans to make wayland default in F25. I'm asking because there was no change proposed for this¹, although Fedora Magazine is stating it will become default in F25².
I know that there have been some considerations on this before³ and some places where issues are collected⁴.
So having been using wayland for the last 5 days on F-25 I have also found the following issues: * gparted won't run - rhbz 1367745 * cut and paste is still weird, but it seems to primarily be in what I think are native wayland apps. EG I can't cut from epiphany/Web and paste into gEdit using either Ctrl/Shit Ins or Ctrl+c/v * window focus seems to be different any non deterministic over X11, if the cursor is over a terminal or text area I sometimes can type straight up sometimes not. Not worked out the exact reproducer for this issue
Peter
On Wed, 2016-08-17 at 15:31 +0100, Peter Robinson wrote:
- cut and paste is still weird, but it seems to primarily be in what
I think are native wayland apps. EG I can't cut from epiphany/Web and paste into gEdit using either Ctrl/Shit Ins or Ctrl+c/v
There is no support for copy/paste in WebKit yet, sorry.
On Wed, Aug 17, 2016 at 3:36 PM, Michael Catanzaro mcatanzaro@gnome.org wrote:
On Wed, 2016-08-17 at 15:31 +0100, Peter Robinson wrote:
- cut and paste is still weird, but it seems to primarily be in what
I think are native wayland apps. EG I can't cut from epiphany/Web and paste into gEdit using either Ctrl/Shit Ins or Ctrl+c/v
There is no support for copy/paste in WebKit yet, sorry.
So why isn't this a blocker for it being default?
On Wed, Aug 17, 2016 at 3:36 PM, Michael Catanzaro mcatanzaro@gnome.org wrote:
On Wed, 2016-08-17 at 15:31 +0100, Peter Robinson wrote:
- cut and paste is still weird, but it seems to primarily be in what
I think are native wayland apps. EG I can't cut from epiphany/Web and paste into gEdit using either Ctrl/Shit Ins or Ctrl+c/v
There is no support for copy/paste in WebKit yet, sorry.
Actually this isn't just WebKit, I can't copy/paste from gEdit into the gnome-shell search either
Peter
On Wed, 2016-08-17 at 19:07 +0100, Peter Robinson wrote:
On Wed, Aug 17, 2016 at 3:36 PM, Michael Catanzaro mcatanzaro@gnome.org wrote:
On Wed, 2016-08-17 at 15:31 +0100, Peter Robinson wrote:
- cut and paste is still weird, but it seems to primarily be in what
I think are native wayland apps. EG I can't cut from epiphany/Web and paste into gEdit using either Ctrl/Shit Ins or Ctrl+c/v
There is no support for copy/paste in WebKit yet, sorry.
Actually this isn't just WebKit, I can't copy/paste from gEdit into the gnome-shell search either
Just to tack on (a bit late) to the copy/paste talk, this bug:
https://bugzilla.gnome.org/show_bug.cgi?id=758958
never seems to get much attention, but it's an extreme inconvenience for anyone who uses a password manager, which is lots of people. I cannot really run Wayland on any of my regular-use systems because of that bug, it's just not practical to manually re-type long randomly- generated passwords. And, of course, if you're in public, it's a huge security risk, because you have to leave the password visible on the screen so you can re-type it into the prompt.
And I'll add this one to the inconvenient category. Only one of the mouse arrows works in the VM, the other works in the host, it's not obvious which is which though, and they don't track at the same rate so it's a lot of iteration to get the VM mouse to click where I need it to.
two mouse arrows appear in the VM window, tracking is all wrong https://bugzilla.redhat.com/show_bug.cgi?id=1370887
On Thu, 2016-09-01 at 12:05 -0700, Adam Williamson wrote:
On Wed, 2016-08-17 at 19:07 +0100, Peter Robinson wrote:
On Wed, Aug 17, 2016 at 3:36 PM, Michael Catanzaro <mcatanzaro@gnom e.org> wrote:
On Wed, 2016-08-17 at 15:31 +0100, Peter Robinson wrote:
- cut and paste is still weird, but it seems to primarily be in
what I think are native wayland apps. EG I can't cut from epiphany/Web and paste into gEdit using either Ctrl/Shit Ins or Ctrl+c/v
There is no support for copy/paste in WebKit yet, sorry.
Actually this isn't just WebKit, I can't copy/paste from gEdit into the gnome-shell search either
Just to tack on (a bit late) to the copy/paste talk, this bug:
https://bugzilla.gnome.org/show_bug.cgi?id=758958
never seems to get much attention, but it's an extreme inconvenience for anyone who uses a password manager, which is lots of people. I cannot really run Wayland on any of my regular-use systems because of that bug, it's just not practical to manually re-type long randomly- generated passwords. And, of course, if you're in public, it's a huge security risk, because you have to leave the password visible on the screen so you can re-type it into the prompt. -- desktop mailing list desktop@lists.fedoraproject.org https://lists.fedoraproject.org/admin/lists/desktop@lists.fedoraproje ct.org
I also depend upon a password manager which I wrote and have used for several years. I began testing it on Wayland in Fedora 24 and found that it failed to paste properly. This week I have tested in Fedora 25 alpha. The good news is after rebuilding it on F25 it works with gedit. The clipboard code is in C connected to python3 with Cython. The bad news is your problem with getting a password into SSH described in bug #758958 fails for me too with code that works properly now with gedit. The shell uses Gtk3 but several of the Gtk3 clipboard functions are not available by introspection through gjs. I know the bugs will be fixed, but copying and pasting is a fundamental operation for the user. If it does not always work it should be fixed before final release.
On Wed, Aug 17, 2016 at 04:36:00PM +0200, Michael Catanzaro wrote:
On Wed, 2016-08-17 at 15:31 +0100, Peter Robinson wrote:
- cut and paste is still weird, but it seems to primarily be in what
I think are native wayland apps. EG I can't cut from epiphany/Web and paste into gEdit using either Ctrl/Shit Ins or Ctrl+c/v
There is no support for copy/paste in WebKit yet, sorry.
That's a reason for not going to Wayland by default in Fedora 25 if I ever saw one. Basic functionality missing.
On Wed, Aug 17, 2016 at 02:10:34PM -0400, Chuck Anderson wrote:
There is no support for copy/paste in WebKit yet, sorry.
That's a reason for not going to Wayland by default in Fedora 25 if I ever saw one. Basic functionality missing.
The problem is that gets into subjective territory *real quick*. There's an attempt to codify features and intended feature parity at https://fedoraproject.org/wiki/Wayland_features. That could use an update with what we've learned from F24 so far.
The top of that page says "But not all features listed here are equally important, there might even be some features listed which will not be implemented in Wayland eventually because they are not suitable." I think Workstation WG needs to prioritize and draw a line in the sand — even if somewhat arbitrary, at least it will be a common point. (And that includes "completion" for each item. For example, there's "copy/paste works everywhere" as a possibility, or else "copy/paste works on all apps installed by default", or any number of other possible lines.)
Hi,
So having been using wayland for the last 5 days on F-25 I have also found the following issues:
- gparted won't run - rhbz 1367745
Seems you filed a duplicate of existing bug 1274451:
https://bugzilla.redhat.com/show_bug.cgi?id=1274451
You could try this as a workaround:
$ xhost +si:localuser:root
prior to run "sudo gparted".
- cut and paste is still weird, but it seems to primarily be in what I
think are native wayland apps. EG I can't cut from epiphany/Web and paste into gEdit using either Ctrl/Shit Ins or Ctrl+c/v
https://bugs.webkit.org/show_bug.cgi?id=146574
- window focus seems to be different any non deterministic over X11,
if the cursor is over a terminal or text area I sometimes can type straight up sometimes not. Not worked out the exact reproducer for this issue
Most likely https://bugzilla.gnome.org/show_bug.cgi?id=756272 fixed in Xwayland upstream (but not part of a released version yet).
Cheers, Olivier.
On 08/17/2016 04:58 PM, Olivier Fourdan wrote:
- window focus seems to be different any non deterministic over X11,
if the cursor is over a terminal or text area I sometimes can type straight up sometimes not. Not worked out the exact reproducer for this issue
Most likely https://bugzilla.gnome.org/show_bug.cgi?id=756272 fixed in Xwayland upstream (but not part of a released version yet).
I backported this and some other XWayland fixes together with Olivier and they should be in https://bodhi.fedoraproject.org/updates/xorg-x11-server-1.18.4-2.fc25 if anyone wants to see if it makes the focus issues better.
I'd like to point out one rather significant thing that came from my conversation with Kamil Paral today: Wayland has not been proposed as a change for Fedora 25 and no one outside the workstation group knows it's still a plan unless they follow our mailing list closely. Fedora QA doesn't test Wayland, Wayland test cases are not part of test matrices and so forth. Now for the rest of the Fedora Project we're going with X11 for F25 and Wayland is still just the experimental thingy. To make it default it needs to go through thorough testing by the QA team and it needs to be part of test matrices.
Jiri
BTW my laptop with F24 now freezes when running on Wayland and undocking. That's a deal breaker for me and I suppose it's an important use case for our target audience. Christian Stadelmann píše v Po 15. 08. 2016 v 14:21 +0000:
I'd like to know whether there are plans to make wayland default in F25. I'm asking because there was no change proposed for this¹, although Fedora Magazine is stating it will become default in F25².
I know that there have been some considerations on this before³ and some places where issues are collected⁴.
Still, the wiki page is missing some important stuff:
How do we handle the fact that even XWayland applications cannot be started as root? I'm ok with dropping the "feature" of starting GUI applications as root, but this might need some discussion and will need some communication. Please note that this would drop any support for at least a dozen applications such as GParted or grub-customizer. And, as Kamil Paral stated on the december thread:
I think we're going to get much worse publicity with this than primary selection. Inexperienced Linux users are used to run "sudo gedit", because terminal editors are unfriendly. Even I, with my years of experience, use "sudo meld" to merge configuration files, because I don't want to learn vimdiff, and graphical apps can simply offer a much superior experience. There was a discussion on devel list about this and everybody talked why running root gui apps is unsafe on X11, but nobody explained why it is unsafe on Wayland. Polkit and fine-grained permissions were proposed, which is a good idea, but it's not going to solve all use cases and it's not the current state of things in many apps. So unless there's a very good reason for deviating from current common practices, I'd put this onto "needs to be fixed" list. The more things we break without a really good reason, the more pushback we're going to get. Maybe this can be improved gradually over time?
This point should at least be noted on the wiki list. Can someone please add it?
Accessibility is pretty much unusable on wayland right now. The wiki page lets me assume that nobody really worked on this besides some minor bug fixings. Can we disable wayland for people relying on a11y? If not, shipping wayland by default probably is a bad idea.
Virtual machine and remote desktop software: Currently, it is impossible for remote desktop and virtual machine GUIs (both under XWayland and Wayland) to grab keys. This is a major regression for all users of these pieces of software.⁵ This point is completely missing on the wiki page. Can someone please add it?
Stability: There are more crashers in gnome-shell with wayland than with X11. That's probably ok, but still worth considering.
Drivers: Lastly, I don't know anything about support with non-Intel GPUs, because their drivers (especially proprietary ones) used to mess up pretty much last time I used them.
¹ https://fedoraproject.org/wiki/Releases/25/ChangeSet
² https://fedoramagazine.org/whats-new-fedora-24-workstation/
³ See e.g:
- "Plan to organize the Wayland-by-default effort" in late 2015, http
s://lists.fedoraproject.org/archives/list/desktop@lists.fedoraproject .org/thread/KL4TW7D263BL56HO7RGEODSYFRSTISYX/
- "Wayland blockers for F24 default" in februar, https://lists.fedora
project.org/archives/list/desktop@lists.fedoraproject.org/thread/FWBG WJALA2D52NZYU3VBUR3G4JCLO56W/
⁴ See https://bugzilla.redhat.com/show_bug.cgi?id=WaylandRelated https://bugzilla.gnome.org/show_bug.cgi?id=WaylandRelated and https://bugzilla.gnome.org/show_bug.cgi?id=wayland https://fedoraproject.org/wiki/Wayland_features
⁵ https://bugs.freedesktop.org/show_bug.cgi?id=97333 and https://bugzilla.redhat.com/show_bug.cgi?id=1285770 -- desktop mailing list desktop@lists.fedoraproject.org https://lists.fedoraproject.org/admin/lists/desktop@lists.fedoraproje ct.org
On Wed, Aug 17, 2016 at 4:51 PM, Jiri Eischmann eischmann@redhat.com wrote:
I'd like to point out one rather significant thing that came from my conversation with Kamil Paral today: Wayland has not been proposed as a change for Fedora 25 and no one outside the workstation group knows it's still a plan unless they follow our mailing list closely. Fedora QA doesn't test Wayland, Wayland test cases are not part of test matrices and so forth. Now for the rest of the Fedora Project we're going with X11 for F25 and Wayland is still just the experimental thingy. To make it default it needs to go through thorough testing by the QA team and it needs to be part of test matrices.
Can someone flip the bits to make X11 the default again on F-25 then, it currently defaults to wayland.
BTW my laptop with F24 now freezes when running on Wayland and undocking. That's a deal breaker for me and I suppose it's an important use case for our target audience.
Christian Stadelmann píše v Po 15. 08. 2016 v 14:21 +0000:
I'd like to know whether there are plans to make wayland default in F25. I'm asking because there was no change proposed for this¹, although Fedora Magazine is stating it will become default in F25².
I know that there have been some considerations on this before³ and some places where issues are collected⁴.
Still, the wiki page is missing some important stuff:
How do we handle the fact that even XWayland applications cannot be started as root? I'm ok with dropping the "feature" of starting GUI applications as root, but this might need some discussion and will need some communication. Please note that this would drop any support for at least a dozen applications such as GParted or grub-customizer. And, as Kamil Paral stated on the december thread:
I think we're going to get much worse publicity with this than primary selection. Inexperienced Linux users are used to run "sudo gedit", because terminal editors are unfriendly. Even I, with my years of experience, use "sudo meld" to merge configuration files, because I don't want to learn vimdiff, and graphical apps can simply offer a much superior experience. There was a discussion on devel list about this and everybody talked why running root gui apps is unsafe on X11, but nobody explained why it is unsafe on Wayland. Polkit and fine-grained permissions were proposed, which is a good idea, but it's not going to solve all use cases and it's not the current state of things in many apps. So unless there's a very good reason for deviating from current common practices, I'd put this onto "needs to be fixed" list. The more things we break without a really good reason, the more pushback we're going to get. Maybe this can be improved gradually over time?
This point should at least be noted on the wiki list. Can someone please add it?
Accessibility is pretty much unusable on wayland right now. The wiki page lets me assume that nobody really worked on this besides some minor bug fixings. Can we disable wayland for people relying on a11y? If not, shipping wayland by default probably is a bad idea.
Virtual machine and remote desktop software: Currently, it is impossible for remote desktop and virtual machine GUIs (both under XWayland and Wayland) to grab keys. This is a major regression for all users of these pieces of software.⁵ This point is completely missing on the wiki page. Can someone please add it?
Stability: There are more crashers in gnome-shell with wayland than with X11. That's probably ok, but still worth considering.
Drivers: Lastly, I don't know anything about support with non-Intel GPUs, because their drivers (especially proprietary ones) used to mess up pretty much last time I used them.
¹ https://fedoraproject.org/wiki/Releases/25/ChangeSet
² https://fedoramagazine.org/whats-new-fedora-24-workstation/
³ See e.g:
- "Plan to organize the Wayland-by-default effort" in late 2015, http
s://lists.fedoraproject.org/archives/list/desktop@lists.fedoraproject .org/thread/KL4TW7D263BL56HO7RGEODSYFRSTISYX/
- "Wayland blockers for F24 default" in februar, https://lists.fedora
project.org/archives/list/desktop@lists.fedoraproject.org/thread/FWBG WJALA2D52NZYU3VBUR3G4JCLO56W/
⁴ See https://bugzilla.redhat.com/show_bug.cgi?id=WaylandRelated https://bugzilla.gnome.org/show_bug.cgi?id=WaylandRelated and https://bugzilla.gnome.org/show_bug.cgi?id=wayland https://fedoraproject.org/wiki/Wayland_features
⁵ https://bugs.freedesktop.org/show_bug.cgi?id=97333 and https://bugzilla.redhat.com/show_bug.cgi?id=1285770 -- desktop mailing list desktop@lists.fedoraproject.org https://lists.fedoraproject.org/admin/lists/desktop@lists.fedoraproje ct.org
-- desktop mailing list desktop@lists.fedoraproject.org https://lists.fedoraproject.org/admin/lists/desktop@lists.fedoraproject.org
On Wed, 2016-08-17 at 17:14 +0100, Peter Robinson wrote:
Can someone flip the bits to make X11 the default again on F-25 then, it currently defaults to wayland.
This issue is on the agenda again for the next Workstation WG meeting (August 31 at 10 AM EDT in #fedora-meeting). Everyone is welcome to attend. I prefer to leave Wayland enabled in the meantime.
On Wed, Aug 17, 2016 at 5:46 PM, Michael Catanzaro mcatanzaro@gnome.org wrote:
On Wed, 2016-08-17 at 17:14 +0100, Peter Robinson wrote:
Can someone flip the bits to make X11 the default again on F-25 then, it currently defaults to wayland.
This issue is on the agenda again for the next Workstation WG meeting (August 31 at 10 AM EDT in #fedora-meeting). Everyone is welcome to attend. I prefer to leave Wayland enabled in the meantime.
We're past feature freeze and are in freeze for Alpha, if it's not an approved change by FESCo and QA aren't testing it I'm not sure what there is to discuss here, it should be reverted now and a freeze exception or blocker requested to get the change landed before we go into Alpha RC phase and waste a bunch of people's time.
Peter
On Wed, Aug 17, 2016 at 05:51:02PM +0200, Jiri Eischmann wrote:
conversation with Kamil Paral today: Wayland has not been proposed as a change for Fedora 25 and no one outside the workstation group knows it's still a plan unless they follow our mailing list closely. Fedora QA doesn't test Wayland, Wayland test cases are not part of test matrices and so forth. Now for the rest of the Fedora Project we're going with X11 for F25 and Wayland is still just the experimental thingy. To make it default it needs to go through thorough testing by the QA team and it needs to be part of test matrices.
We really _do_ need to go through the proper process here. Please remember, Fedora is a huge project with hundreds of active contributors, and this isn't just mere bureaucracy. We require the coordination in order to continue to produce the reoccuring miracle that is a functioning Fedora release.
I'm excited as anyone for the new features and improvements Wayland brings to the table, but let's do it right.
On Wed, Aug 17, 2016 at 11:11 AM, Matthew Miller mattdm@fedoraproject.org wrote:
On Wed, Aug 17, 2016 at 05:51:02PM +0200, Jiri Eischmann wrote:
conversation with Kamil Paral today: Wayland has not been proposed as a change for Fedora 25 and no one outside the workstation group knows it's still a plan unless they follow our mailing list closely. Fedora QA doesn't test Wayland, Wayland test cases are not part of test matrices and so forth. Now for the rest of the Fedora Project we're going with X11 for F25 and Wayland is still just the experimental thingy. To make it default it needs to go through thorough testing by the QA team and it needs to be part of test matrices.
We really _do_ need to go through the proper process here. Please remember, Fedora is a huge project with hundreds of active contributors, and this isn't just mere bureaucracy. We require the coordination in order to continue to produce the reoccuring miracle that is a functioning Fedora release.
I'm excited as anyone for the new features and improvements Wayland brings to the table, but let's do it right.
If someone wants to file a FESCo ticket to get it approved as a late change, I think that's reasonable. But tomorrow is go/no-go and I think it's at least as risky to flip back to Xorg by default as it is to leave Wayland the default; and no doubt FESCo would take QA's opinion on the late change into account but this lateness I think is pretty minor compared to some of the late changes that have happened in the past.
I think a case can be made that it was always intended to be the default for Fedora 25, it very nearly was the default for Fedora 24. It's something of an oversight there was no change filed for Fedora 25, and it just slipped through the crack. Unless testers are manually changing to Xorg, it is being tested since it's the enabled default, which is a requirement by the change process. It probably also is at or nearly at the 100% code complete point well before that deadline; and if it's not 100% then the WG can estimate how far away it is and how likely it'd be at 100% by that deadline.
Top on my list of blocking behaviors for which I'm not aware of an appropriate release criterion is: by beta the switching between wayland and X needs to be bulletproof; in particular the ability to switch from Wayland to X must actually work and must stick through a reboot (persistence). As long as the user can reliably use X, I think the worst of Waylands maturation problems are surmountable. There is an in place fallback, it's not like pretty much all other system wide complex changes where there is no such user initiated fallback available.
On Wed, Aug 17, 2016 at 12:02:09PM -0600, Chris Murphy wrote:
On Wed, Aug 17, 2016 at 11:11 AM, Matthew Miller mattdm@fedoraproject.org wrote:
On Wed, Aug 17, 2016 at 05:51:02PM +0200, Jiri Eischmann wrote:
conversation with Kamil Paral today: Wayland has not been proposed as a change for Fedora 25 and no one outside the workstation group knows it's still a plan unless they follow our mailing list closely. Fedora QA doesn't test Wayland, Wayland test cases are not part of test matrices and so forth. Now for the rest of the Fedora Project we're going with X11 for F25 and Wayland is still just the experimental thingy. To make it default it needs to go through thorough testing by the QA team and it needs to be part of test matrices.
We really _do_ need to go through the proper process here. Please remember, Fedora is a huge project with hundreds of active contributors, and this isn't just mere bureaucracy. We require the coordination in order to continue to produce the reoccuring miracle that is a functioning Fedora release.
I'm excited as anyone for the new features and improvements Wayland brings to the table, but let's do it right.
If someone wants to file a FESCo ticket to get it approved as a late change, I think that's reasonable. But tomorrow is go/no-go and I think it's at least as risky to flip back to Xorg by default as it is to leave Wayland the default; and no doubt FESCo would take QA's opinion on the late change into account but this lateness I think is pretty minor compared to some of the late changes that have happened in the past.
I think a case can be made that it was always intended to be the default for Fedora 25, it very nearly was the default for Fedora 24. It's something of an oversight there was no change filed for Fedora 25, and it just slipped through the crack. Unless testers are manually changing to Xorg, it is being tested since it's the enabled default, which is a requirement by the change process. It probably also is at or nearly at the 100% code complete point well before that deadline; and if it's not 100% then the WG can estimate how far away it is and how likely it'd be at 100% by that deadline.
Top on my list of blocking behaviors for which I'm not aware of an appropriate release criterion is: by beta the switching between wayland and X needs to be bulletproof; in particular the ability to switch from Wayland to X must actually work and must stick through a reboot (persistence). As long as the user can reliably use X, I think the worst of Waylands maturation problems are surmountable. There is an in place fallback, it's not like pretty much all other system wide complex changes where there is no such user initiated fallback available.
I don't think that this is an accurate description of current situation. If testers who install F25 get wayland by default, all this tells us that it mostly works, but not that it's ready for widespread use. There's still a long list of open issues until feature parity with X11. Those things don't get reported as bugs, because they are well known missing features, not really bugs, but they cannot be ignored. I've been using gnome-wayland myself for the last year, and I think it's great, but there's still too many shortcomings.
IMO, the only reasonable course of action at this point is to make X11 the default for F25 and punt wayland on to F26.
Zbyszek
On Thu, Aug 18, 2016 at 02:32:34AM +0000, Zbigniew Jędrzejewski-Szmek wrote:
On Wed, Aug 17, 2016 at 12:02:09PM -0600, Chris Murphy wrote:
On Wed, Aug 17, 2016 at 11:11 AM, Matthew Miller mattdm@fedoraproject.org wrote:
On Wed, Aug 17, 2016 at 05:51:02PM +0200, Jiri Eischmann wrote:
conversation with Kamil Paral today: Wayland has not been proposed as a change for Fedora 25 and no one outside the workstation group knows it's still a plan unless they follow our mailing list closely. Fedora QA doesn't test Wayland, Wayland test cases are not part of test matrices and so forth. Now for the rest of the Fedora Project we're going with X11 for F25 and Wayland is still just the experimental thingy. To make it default it needs to go through thorough testing by the QA team and it needs to be part of test matrices.
We really _do_ need to go through the proper process here. Please remember, Fedora is a huge project with hundreds of active contributors, and this isn't just mere bureaucracy. We require the coordination in order to continue to produce the reoccuring miracle that is a functioning Fedora release.
I'm excited as anyone for the new features and improvements Wayland brings to the table, but let's do it right.
If someone wants to file a FESCo ticket to get it approved as a late change, I think that's reasonable. But tomorrow is go/no-go and I think it's at least as risky to flip back to Xorg by default as it is to leave Wayland the default; and no doubt FESCo would take QA's opinion on the late change into account but this lateness I think is pretty minor compared to some of the late changes that have happened in the past.
I think a case can be made that it was always intended to be the default for Fedora 25, it very nearly was the default for Fedora 24. It's something of an oversight there was no change filed for Fedora 25, and it just slipped through the crack. Unless testers are manually changing to Xorg, it is being tested since it's the enabled default, which is a requirement by the change process. It probably also is at or nearly at the 100% code complete point well before that deadline; and if it's not 100% then the WG can estimate how far away it is and how likely it'd be at 100% by that deadline.
Top on my list of blocking behaviors for which I'm not aware of an appropriate release criterion is: by beta the switching between wayland and X needs to be bulletproof; in particular the ability to switch from Wayland to X must actually work and must stick through a reboot (persistence). As long as the user can reliably use X, I think the worst of Waylands maturation problems are surmountable. There is an in place fallback, it's not like pretty much all other system wide complex changes where there is no such user initiated fallback available.
I don't think that this is an accurate description of current situation. If testers who install F25 get wayland by default, all this tells us that it mostly works, but not that it's ready for widespread use. There's still a long list of open issues until feature parity with X11. Those things don't get reported as bugs, because they are well known missing features, not really bugs, but they cannot be ignored. I've been using gnome-wayland myself for the last year, and I think it's great, but there's still too many shortcomings.
IMO, the only reasonable course of action at this point is to make X11 the default for F25 and punt wayland on to F26.
Is "feature parity with X11" the intended measure of readiness here? I thought there were specific X11 features that are intentionally not going to be duplicated in Wayland, which would make that a false bar to try to reach.
We do have this existing feature page: https://fedoraproject.org/wiki/Changes/WaylandByDefault
I was asked to propose this as a late feature for an exception. However, the page doesn't appear to be updated technically. Also, this test criterion seems suspect:
"Use the desktop normally, and verify that there are no obvious instabilities, or Wayland-specific bugs or performance problems"
"No Wayland-specific bugs" also doesn't appear to jibe with the desire to get Wayland out as a default even if there are a few specific bugs to solve. The page lists https://bugzilla.redhat.com/1277927 as a tracker, but I didn't think the objective is to fix/close all the bugs on that tracker in order to move to Wayland.
On Thu, Aug 18, 2016 at 03:07:19PM -0400, Paul W. Frields wrote:
Is "feature parity with X11" the intended measure of readiness here? I thought there were specific X11 features that are intentionally not going to be duplicated in Wayland, which would make that a false bar to try to reach.
We've been kind of selling it as "it's ready when most people won't notice". But I think we should nail down a precise bar. (Hmmm. Mixed metaphor.)
We do have this existing feature page: https://fedoraproject.org/wiki/Changes/WaylandByDefault
Yeah — and that's a really great, useful thing. I'd _really_ love if there were a "good enough!" criterion listed for each feature, agreed on formally by the WG.
[...]
"Use the desktop normally, and verify that there are no obvious instabilities, or Wayland-specific bugs or performance problems" "No Wayland-specific bugs" also doesn't appear to jibe with the desire
Maybe "No Wayland-specific bugs which would qualify as blockers on their own", or something?
On Wed, Aug 17, 2016 at 12:02 PM, Chris Murphy lists@colorremedies.com wrote:
On Wed, Aug 17, 2016 at 11:11 AM, Matthew Miller mattdm@fedoraproject.org wrote:
On Wed, Aug 17, 2016 at 05:51:02PM +0200, Jiri Eischmann wrote:
conversation with Kamil Paral today: Wayland has not been proposed as a change for Fedora 25 and no one outside the workstation group knows it's still a plan unless they follow our mailing list closely. Fedora QA doesn't test Wayland, Wayland test cases are not part of test matrices and so forth. Now for the rest of the Fedora Project we're going with X11 for F25 and Wayland is still just the experimental thingy. To make it default it needs to go through thorough testing by the QA team and it needs to be part of test matrices.
We really _do_ need to go through the proper process here. Please remember, Fedora is a huge project with hundreds of active contributors, and this isn't just mere bureaucracy. We require the coordination in order to continue to produce the reoccuring miracle that is a functioning Fedora release.
I'm excited as anyone for the new features and improvements Wayland brings to the table, but let's do it right.
If someone wants to file a FESCo ticket to get it approved as a late change, I think that's reasonable. But tomorrow is go/no-go and I think it's at least as risky to flip back to Xorg by default as it is to leave Wayland the default; and no doubt FESCo would take QA's opinion on the late change into account but this lateness I think is pretty minor compared to some of the late changes that have happened in the past.
Since it's no go, there's a good chance filing a ticket sometime today means it gets reviewed by FESCo at this Friday's meeting. Ideally the WG would not have deferred the Wayland question to next week's WG meeting. But a FESCo ticket could at least resolve whether the change would be accepted should the WG decide next week to proceed with Wayland by default; or if it's simply too late now and must be postponed to Fedora 26.
I'm with sgallagh on this, rip off the bandaid. But that's not one of the official process options.
On Thu, Aug 18, 2016 at 12:57:09PM -0600, Chris Murphy wrote:
I'm with sgallagh on this, rip off the bandaid.
I'm coming down on this side too. I think it's going to be somewhat painful, but, honestly, we've had a series of really excellent releases without much excitement, and I think we can afford a little — especially if the X fallback is functional and clearly documented.
But that's not one of the official process options.
+1
On Thu, Aug 18, 2016 at 03:02:22PM -0400, Matthew Miller wrote:
On Thu, Aug 18, 2016 at 12:57:09PM -0600, Chris Murphy wrote:
I'm with sgallagh on this, rip off the bandaid.
I'm coming down on this side too. I think it's going to be somewhat painful, but, honestly, we've had a series of really excellent releases without much excitement, and I think we can afford a little — especially if the X fallback is functional and clearly documented.
But that's not one of the official process options.
+1
I filed https://fedorahosted.org/fesco/ticket/1615 for this purpose.
On Thu, 2016-08-18 at 17:03 -0400, Paul W. Frields wrote:
I filed https://fedorahosted.org/fesco/ticket/1615%C2%A0for this purpose.
We have received approval from FESCo at today's meeting to proceed with Wayland by default. We can continue to discuss the wisdom in doing so, but this should hopefully resolve the concerns about policy/procedure.
Michael
desktop@lists.fedoraproject.org