Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

classic Classic list List threaded Threaded
15 messages Options
Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Simon McVittie-7
Package: gnome
Version: 1:3.30+1
Severity: important
X-Debbugs-Cc: Jonathan Dowland <[hidden email]>, [hidden email], [hidden email]

On the debian-devel mailing list, Jonathan Dowland wrote:

> I was surprised to learn — by way of synaptic being autoremoved — that
> the default desktop in Buster will be GNOME/Wayland. I personally do not
> think that Wayland is a sensible choice for the default *yet*; and if
> the consequence is that bugs for software that do not work properly with
> Wayland have their severity inflated such that they are autoremoved (and
> thus potentially removed entirely from Buster), a decision that — in
> isolation — makes sense to me; although Synaptic is quite a high profile
> package within Debian for this to happen.
>
> I think the default should be reconsidered.

I am not exactly a core member of the team (I joined to help to maintain
GLib), and I didn't make the decision to stop diverging from upstream
on this particular point, so I haven't been able to articulate why that
decision was made.

I can completely understand if other GNOME team members don't want to
interact with the -devel thread, given the amount of negativity it is
attracting from people who don't like GNOME and wouldn't use it whatever
we did; but I think it would be good to reassess the costs and benefits
of Wayland vs. X11 by default, and either make a positive decision to
keep Wayland as the default, or diverge from upstream and switch back
to Xorg by default like Debian stretch and Ubuntu 18.04 did.

We are now in hard freeze, so if the team does want to go back to Xorg
as the default, this is our last chance.

(As the maintainer of flatpak in Debian, I would be sad to see us go
back to a display protocol where every app can copy other apps' window
contents, inject fake input events or be a keylogger... but I'm also
aware that some programs rely on X11 not providing meaningful privilege
separation, and that the Wayland compositor model exacerbates any bugs
that cause GNOME Shell to crash.)

    smcv

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Jonathan Dowland
severity 927667 serious
thanks

[ I wrote the following reply to smcv's last mail to this bug, but only
sent it to -desktop: https://lists.debian.org/debian-desktop/2019/05/msg00000.html ]

Thanks to Simon for filing this bug in the first place and for engaging
in discussion about the issue.

My personal conclusion is that we, as a project, need to spend more time
carefully evaluating Wayland prior to it being the default desktop technology
for Debian, following the approach we have used for other major technologies
adopted by default (e.g. systemd). This hasn't happened for Wayland yet, and
I'm concerned that it may just not be quite ready (e.g. #928030, #928002).

I want an explicit ACK or NACK on this, and so I'm bumping severity so that
this bug can't languish past Buster release. Justification is the regression
in behavour wrt X11 (as documented in those two bugs amongst others) break
unrelated software. I'll interpret downgrading as being a NACK, but I would
*really* appreciate engagement instead of just bug severity changes.

Two ways of resolving this are: Either the default GNOME3 session in Debian
switched back to Xorg, or the default desktop session is switched away from
GNOME; but I would much prefer the former.


--
Jonathan Dowland

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Jonathan Dowland
In reply to this post by Simon McVittie-7

Hi Michael,

Thank you for sharing your take.

On Sat, May 11, 2019 at 11:21:22AM +0200, Michael Biebl wrote:
>Then there is the case, that I sometimes need to use software like
>teamviewer (I know, bad proprietary software), which is not Wayland
>ready. I need something cross-plattform though, and I'm not aware of
>another solution which runs on top of Wayland.

I think we should have some sympathy towards users of non-free software
on top of Debian, even if we aren't packaging it. Pragmatically many of
our users rely on it or want it. E.g. NVIDIA driver users, Google Chrome,
Steam… and IMHO we further our goals if we are sensitive to their needs
rather than making life harder for them.

>Incidentally, I was the one responsible for switching back the default
>to Xorg in stretch and the main reasons which guided my decision back
>then mostly still apply today.

I thought you might have been, I started reading the git log of some
packages to try and figure out what patch would need to be written, I
saw your name in (I think) gdm3 (I haven't yet come up with a working
patch but this gave me some clues thank you)

>That said, I'm no longer an active GNOME team member and haven't really
>done any GNOME related work during the buster development cycle.

I hadn't realised you were no longer active in the GNOME team. I hope
you are still active in Debian :-) I've always enjoyed interacting with
you (and you may not remember but I think we met at DebConf '07)

>And I'm convinced, that the people doing the work should decide.

Mostly agree :-)

--

⢀⣴⠾⠻⢶⣦⠀
⣾⠁⢠⠒⠀⣿⡁ Jonathan Dowland
⢿⡄⠘⠷⠚⠋⠀ https://jmtd.net
⠈⠳⣄⠀⠀⠀⠀

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Jonathan Dowland
In reply to this post by Jonathan Dowland
reassign 927667 gdm3
tags 927667 +patch
thanks

On Fri, May 10, 2019 at 10:26:07AM +0100, Jonathan Dowland wrote:
>Two ways of resolving this are: Either the default GNOME3 session in Debian
>switched back to Xorg, or the default desktop session is switched away from
>GNOME; but I would much prefer the former.

The attached debdiff (bringing back the similar change from 2016) seems
to work for me (tested in a VM) but I would appreciate any additional
testing that folks could spare.

The debdiff is an NMU but I currently have no plans to upload it as such.

Also available at:
    https://salsa.debian.org/gnome-team/gdm/merge_requests/8

--

⢀⣴⠾⠻⢶⣦⠀
⣾⠁⢠⠒⠀⣿⡁ Jonathan Dowland
⢿⡄⠘⠷⠚⠋⠀ https://jmtd.net
⠈⠳⣄⠀⠀⠀⠀

927667_default_session.patch (2K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Samuel Thibault-8
Hello,

Just so that the info is recorded here as well: as discussed on

https://lists.debian.org/debian-accessibility/2019/02/msg00004.html

Wayland does not provide yet the support that the Orca screen reader
needs to provide proper support for blind people (keyboard shortcuts for
GUI review, basically).

Samuel

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Jonathan Dowland
In reply to this post by Simon McVittie-7
Hi Andrew

So far as I am aware there is still radio silence from active GNOME
packaging team members regarding this issue. No comment yet on the
patch I adapted, positive or negative; this bug (#927667) remains
at an RC severity.

I've not yet read all the thread that Samuel linked to[1] but it looks
like it leans in favour of preserving the current default (xorg).

I'm copying -release team to see if they have any (new) opinions on
the matter. Otherwise I guess it's up to someone to prepare an NMU
upload, which I will *try* to look at in the next few days, but can't
make any guarantees.

Further testers of the patch on this bug would be most welcome.

[1] https://lists.debian.org/debian-accessibility/2019/02/threads.html#00004

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Ansgar Burchardt-5
Jonathan Dowland writes:

> So far as I am aware there is still radio silence from active GNOME
> packaging team members regarding this issue. No comment yet on the
> patch I adapted, positive or negative; this bug (#927667) remains
> at an RC severity.
>
> I've not yet read all the thread that Samuel linked to[1] but it looks
> like it leans in favour of preserving the current default (xorg).
>
> I'm copying -release team to see if they have any (new) opinions on
> the matter. Otherwise I guess it's up to someone to prepare an NMU
> upload, which I will *try* to look at in the next few days, but can't
> make any guarantees.

I'm just a GNOME user, but from gdm3's changelog the default was
switched to Wayland in July 2017 (or August 2017 for unstable).  I
myself only noticed the switch after reading it happened somewhere on
the internet shortly after it happened.

Switching the default back two weeks before the release seems too late
for me.  The largest issue seems to be accessibility, but as far as I
understand we already recommend a different desktop environment for
that.  I don't think that warrants changes that would only see very
little testing by now :-/

Ansgar

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Simon McVittie-7
In reply to this post by Jonathan Dowland
On Wed, 19 Jun 2019 at 17:33:55 +0100, Jonathan Dowland wrote:
> So far as I am aware there is still radio silence from active GNOME
> packaging team members regarding this issue. No comment yet on the
> patch I adapted, positive or negative; this bug (#927667) remains
> at an RC severity.

(Adding debian-gtk-gnome, debian-desktop and some people who might have
useful input to Cc)

In case anyone else in the GNOME team has got the wrong idea from my
involvement in #927667: I don't think I am the right person to make a
decision on this. So if GNOME team members are waiting for me to either
make an upload changing the default back to X11, or veto such an upload:
don't wait for that, it is unlikely to happen (unless the team cannot
make a decision and punts this to the technical committee, but I hope
we don't have to resort to that).

I would very much appreciate input from the rest of the team, particularly:

- Laurent: I know you've had strong opinions about using Wayland for GNOME.
  Do you feel strongly that Debian should be defaulting to Wayland? Are
  there any reasons for that default that are missing from my attempt to
  summarize earlier on the bug?
- Ubuntu GNOME team: which recent Ubuntu versions, if any, are using
  Wayland for their GNOME-based desktop?

I've left some comments on
https://salsa.debian.org/gnome-team/gdm/merge_requests/8 regarding the
technical side of the proposed change.

Thanks,
    smcv

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Sebastien Bacher-2
Hey there,

Le 19/06/2019 à 22:19, Simon McVittie a écrit :
> - Ubuntu GNOME team: which recent Ubuntu versions, if any, are using
>    Wayland for their GNOME-based desktop?

We don't have any supported Ubuntu version using Wayland by default, our
motivations for sticking to Xorg were mostly desktop sharing/rdp support
and the fact that under wayland a gnome-shell segfault takes the whole
session down without giving user a chance to save their work.
While screen sharing is being actively being worked on, our metrics show
that gnome-shell errors are still quite common, even in recent versions
so it's not likely that we change our default for the next LTS.

Cheers,
Sebastien Bacher

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Ansgar Burchardt-5
In reply to this post by Simon McVittie-7
Michael Biebl writes:
> On Wed, 19 Jun 2019 22:16:58 +0200 Ansgar Burchardt <[hidden email]> wrote:
>> I'm just a GNOME user, but from gdm3's changelog the default was
>> switched to Wayland in July 2017 (or August 2017 for unstable).  I
>> myself only noticed the switch after reading it happened somewhere on
>> the internet shortly after it happened.
[...]
> When you talk about "gdm3's default", what exactly do you mean: the
> display technology gdm is using or the type of GNOME session that is
> started as default?

I meant the GNOME session; that is the only thing that changed in
July/August 2017 as gdm itself used Wayland already before as far as I
understand.

For something that has been the default for almost two years to be
switched back two weeks before the release there has to be some really
huge issues for me.  Any change will not see much testing given the time
after all and risks new regressions.

Ansgar

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Iain Lane-8
In reply to this post by Simon McVittie-7
Hi Simon, others,

Thanks to those involved so far in shepherding this side of the issue
along. And also for poking me (repeatedly!) to reply.

On Wed, Jun 19, 2019 at 09:19:49PM +0100, Simon McVittie wrote:
> On Wed, 19 Jun 2019 at 17:33:55 +0100, Jonathan Dowland wrote:
> > So far as I am aware there is still radio silence from active GNOME
> > packaging team members regarding this issue. No comment yet on the
> > patch I adapted, positive or negative; this bug (#927667) remains
> > at an RC severity.

Right, sorry about that. It's a big lack of spoons for me (please
consider that if you are thinking about rebutting my points below). I
personally find this kind of topic a bit draining, although I appreciate
that the temperature of the bug (not the previous -devel thread) is
quite low.  Thanks for that.

> (Adding debian-gtk-gnome, debian-desktop and some people who might have
> useful input to Cc)
> […]
> - Ubuntu GNOME team: which recent Ubuntu versions, if any, are using
>   Wayland for their GNOME-based desktop?

Seb's explained that, and the reasons why that decision was made. So you
can have my inconclusive personal opinion:

I've been deliberately using Wayland-on-Ubuntu and Wayland-on-Debian the
whole time, and the only thing that has irritated *me* was one time I
couldn't screen share from Firefox. I *like* that it's raised the bar
for privilege separation (the Synaptic case), as I find it quite
distasteful to be running the entire thing as root. I acknowledge that
the shell-crashing-crashes-the-session model could be uncomfortable
sometimes — but my experience has not been one where that happens. Some
of the upstream improvements (e.g. fractional scaling, which is still
experimental) are Wayland-only.

I've read and understood the counter-arguments posted on the bug. I
don't feel in a particularly good position with respect to weighing up
the balance here. If you are affected by a Wayland-specific bug,
especially if it impacts you frequently (e.g. you can't do something you
need to do, or the shell repeatedly crashes / the mouse becomes
unresponsive / similar), that is going to be really quite irritating.

And I share Ansgar's concern about this all being very late now. Part of
that is down to our (GNOME team at large) lack of engagement on the bug
— apologies again — but it would have felt late to me even at the end of
April. That said, Ubuntu has been shipping with this configuration and
we don't know of any particular issues. Release team, what do you think?

In conclusion, I will not stand in the way of anyone if they want to
execute this change, but it's not likely to be me doing it.

> I've left some comments on
> https://salsa.debian.org/gnome-team/gdm/merge_requests/8 regarding the
> technical side of the proposed change.

Someone could probably look in Ubuntu's gdm3 package to see what we're
doing. We provide "GNOME" (Xorg, the default) and "GNOME on Wayland"
sessions.

Cheers,

--
Iain Lane                                  [ [hidden email] ]
Debian Developer                                   [ [hidden email] ]
Ubuntu Developer                                   [ [hidden email] ]

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Michael Biebl-3
Hi everyone

Am 20.06.19 um 11:12 schrieb Iain Lane:
>> I've left some comments on
>> https://salsa.debian.org/gnome-team/gdm/merge_requests/8 regarding the
>> technical side of the proposed change.
> Someone could probably look in Ubuntu's gdm3 package to see what we're
> doing. We provide "GNOME" (Xorg, the default) and "GNOME on Wayland"
> sessions.

Afair, this required changing gnome-session. I left a comment in the gdm MR.
If the point is, to not switch the desktop session automatically on
upgrades, then the session files would have to be renamed (back again)
to gnome.desktop (Xorg) and gnome-wayland.desktop from gnome.desktop
(Wayland) and gnome-xorg.desktop.
At least this is how I remember the details from back then in 2016.
I haven't checked if the situation is still the same today.

Regards,
Michael

--
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?


signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Sam Hartman-5
In reply to this post by Iain Lane-8

I'm writing with my DPL hat on in the role of a facilitator/mediator.
I have no actual power in this situation and it is entirely reasonable
to ignore me.

I feel very uncomfortable with a change as big as this revert happening
this late in the release cycle.
I think that my reading of how the release team handles issues is
sufficient to say that they almost certainly have serious concerns about
that big of a change this late too.

And yet, the lack of a clear reconfirmation in this time line even given
the wonderfully civil discussion is telling.

My proposal--which again I have no power to implement--is that we go
forward with the current default.  However, we remain open to a revert
in the first couple of buster point releases.  The criteria for that
revert should be based on the actual severity and frequence of problems
our users run into, but should specifically exclude the blanket
reluctance to  make a change like that in a point release.
We would still need adequate testing of such a revert.

So, what I think this would require to be a viable proposal is:

* an ack from the buster stable release managers that if we run into
  real problems they would accept such a change

* an ack from gnome that if we do need to do a revert we'd be willing to
  revert in unstable and testing for a while to do as much testing of
  the revert in those environments.  Obviously such testing is imperfect
  given that gnome may (hopefully will) have moved on in Debian by then.

Again, feel free to ignore this message entirely if it does not move the
conversation forward.
I'm just seeing a stuck issue and proposing a way to try and unstick it.

--Sam

signature.asc (497 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Jonathan Dowland
Preserving the rather large CC list for now…

On Thu, Jun 20, 2019 at 06:05:22AM -0400, Sam Hartman wrote:
>I feel very uncomfortable with a change as big as this revert happening
>this late in the release cycle.

How big is big? The MR I raised resurrects a patch that changes one line
of code, and was shipped in the last stable release. Although it looks
like further work is needed to make the change as smooth as possible so
this would grow. However, the patch certainly needs more testing. Is the
issue that it results in a large change in terms of what software is
executed by the user as a consequence, and what of that has been tested
thus far in the freeze?

How late is late? How would you have felt about it back in early April,
when I first raised it? I was surprised to discover it then, and I felt
it was late in the cycle even then. But mostly whats happened since is
nothing.

I absolutely do not blame the GNOME team here. Simon McVittie and
Michael Biebl in particular have taken risks sticking their heads above
the parapet to engage with me on this matter, and both have made it
clear that it would be unreasonable for them to make the call given
their respective levels of involvement. I respect that, and I am
extremely grateful for them engaging with the issue. The others are
either too busy or have taken a decision not to engage with a
potentially toxic issue, and I respect that, too: we are all volunteers
who have to make our own choices about what we are prepared to do and
engage in. Besides, like many teams, the GNOME team is clearly
under-resourced.

I am a *little* disappointed that this does not seem to have been
thought of as an important, project-wide issue. Regardless of whether
one uses GNOME or Wayland oneself, the matter of the default desktop for
the distribution we are all working to produce, and the experience that
our users will get out of the box, I would have thought was important
for all of us. It reinforces the idea, to me, that we are largely
working in our own silos, and not concerned (enough) about the holistic
distribution as a whole.

>And yet, the lack of a clear reconfirmation in this time line even given
>the wonderfully civil discussion is telling.

I'm very pleased that the discussion has come across as civil. I've
tried really hard from my end to achieve that, I know that issues around
GNOME can result in some very toxic communications.

>My proposal--which again I have no power to implement--is that we go
>forward with the current default.  However, we remain open to a revert
>in the first couple of buster point releases.

There are caveats with switching the default in either direction.
Let's say we go with Wayland now, and later decide to switch as per the
criteria/process you sketch below.

 • users of the default, who got Wayland from Buster onwards and had no
   problems, would subsequently find themselves switched to Xorg by
   stable-updates, which IMHO would be unexpected (if noticed) and
   contrary to the expectations of a stable release.

 • A user who installed or upgraded and got Wayland by default but had
   problems, would have likely addressed them by switching to the Xorg
   session explicitly (assuming they could figure out that doing so
   mitigated their issues). Changing the default would only prevent
   *future* users from hitting the same problems.

>The criteria for that revert should be based on the actual severity and
>frequence of problems our users run into, but should specifically
>exclude the blanket reluctance to  make a change like that in a point
>release.  We would still need adequate testing of such a revert.

My concern with this is it's a new set of policies and procedures, not
codified anywhere, with a lot of detail to work out "on the hoof" (how
do we measure frequency of problems? do we go with the existing bug
severity guidelines? How much is adequate testing? etc.)

So combined with the user experience above, I think we would be best not
to change the default within a stable release cycle, unless there was
some kind of enormous catastrophic issue with Wayland that we don't
know about yet, and that's unlikely.

I still argue that the traditional Debian conservative, when-its-ready
approach would be the distribution status quo (Xorg), but I recognise
the concerns about the proposed patch, further work needed, lack of
testing etc.; and those are not issues I think I can resolve alone.


--
Jonathan Dowland

Reply | Threaded
Open this post in threaded view
|

Bug#927667: gnome: please confirm or revert choice of Wayland for default desktop

Laurent Bigonville-5
In reply to this post by Simon McVittie-7
On 19/06/19 22:19, Simon McVittie wrote:
> [...]
> I would very much appreciate input from the rest of the team, particularly:
>
> - Laurent: I know you've had strong opinions about using Wayland for GNOME.
>    Do you feel strongly that Debian should be defaulting to Wayland? Are
>    there any reasons for that default that are missing from my attempt to
>    summarize earlier on the bug?
> [...]

I'm personally using wayland for more than 3 years on my work laptop
(Intel card) and my home desktop (ATI/AMD with OSS driver) and even if
there were transient issues at some point, everything is pretty stable
now with 3.30 (the version that will be released with buster). Like Iain
the main annoyance I have in my daily use is with the desktop/window
sharing in firefox.

Wayland has been (re)made the default in debian back in July 2017
(beginning of the dev cycle for buster), I don't remember receiving any
objections at the time. The question about using it by default was
raised by Jonathan in Apr 2019, two months in the (soft) freeze, it was
already quite late at that point IMHO to switch back. This makes me
wonder, are there even people using GNOME in sid/testing? Are there
people testing with the default settings or has everybody switch back to
X11? Because we had a full development cycle and we didn't have a
massive number of bugs being filled about this, how should we interpret
that?

It's also important to note that we are not pioneer in this, Fedora is
defaulting to GNOME Wayland since Fedora 25 (Nov 2016). Both RHEL 8
(just released and using GNOME 3.28, so one release lower) and SUSE
Linux Enterprise Desktop 15 (released in end of June last year using
GNOME 3.26) are also defaulting to GNOME Wayland.

We could indeed revert to X11 in a point release if things are going
horribly wrong, some first step could be to put more information about
this in the release notes. RHEL has
https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html-single/8.0_release_notes/index#desktop 
but I don't think that everything there applies to Debian