Bug 30041 - Evolution & Epiphany windows does not appear after webkit2 upgrade
Summary: Evolution & Epiphany windows does not appear after webkit2 upgrade
Status: RESOLVED FIXED
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: 8
Hardware: All Linux
Priority: High critical
Target Milestone: ---
Assignee: All Packagers
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on: 30064
Blocks:
  Show dependency treegraph
 
Reported: 2022-02-13 09:42 CET by aguador
Modified: 2022-02-18 16:18 CET (History)
9 users (show)

See Also:
Source RPM: webkit2-2.34.5-1.mga8.src.rpm
CVE:
Status comment: Workaround: Downgrade, see c15


Attachments
urpmi command to test if all packages used by epiphany are installed (3.03 KB, text/plain)
2022-02-14 17:56 CET, Dave Hodgins
Details

Description aguador 2022-02-13 09:42:57 CET
After the latest upgrade of webkit2 Evolution does not draw a window even though the processes are running in background, presumably the reason no error message appears.

This has been tested after restart, with both light and dark themes and in both Enlightenment and IceWM with the same failure.
Comment 1 aguador 2022-02-13 10:18:55 CET
I can confirm that downgrading to webkit2-2.34.4 fixes the problem.
Comment 2 Morgan Leijström 2022-02-13 12:43:07 CET
Ouch, that is bad.

Thank you for the heads up and testing downgrading.

No specified packager, assigning all.

If you start evolution in a terminal, do you see some interesting output?

Anything in the journal?

Severity: normal => critical
Priority: Normal => High
Assignee: bugsquad => pkg-bugs
CC: (none) => fri

Comment 3 Cyril Levet 2022-02-13 13:43:43 CET
Hi,
I confirm. Terminal output is :

(evolution-alarm-notify:2475): GLib-GIO-WARNING **: 13:33:38.982: Your application did not unregister from D-Bus before destruction. Consider using g_application_run().
Oops, secure memory pool already initialized
WaylandCompositor requires eglBindWaylandDisplayWL, eglUnbindWaylandDisplayWL and eglQueryWaylandBuffer.
Nested Wayland compositor could not initialize EGL
bwrap: Can't create file at /run/user/1000/bus: No such device or address

The journal does not show anything.

CC: (none) => cyril.levet0780

Comment 4 Thomas Backlund 2022-02-13 13:53:25 CET
please try if  evolution-3.38.3-3.mga8 in updates testing works any better
Comment 5 sturmvogel 2022-02-13 15:29:05 CET
evolution-3.38.3-3.mga8 under a Plasma 64bit VM doesn't solve the problem. The output only got shorter:

Oops, secure memory pool already initialized
bwrap: Can't create file at /run/user/1000/bus: No such device or address
Comment 6 Thomas Andrews 2022-02-13 15:34:30 CET
I'm not an Evolution user, but I do have a MGA8 Gnome guest in VirtualBox.

Confirmed that the Evolution window would open before the webkit2 update, and won't open after. I updated using qarepo to get just the evolution package, ran it again from the command line, with no joy. This is what it said in Terminal, different from what was reported in Comment 3:


(evolution-alarm-notify:2515): GLib-GIO-WARNING **: 09:28:00.091: Your application did not unregister from D-Bus before destruction. Consider using g_application_run().
Oops, secure memory pool already initialized
bwrap: Can't create file at /run/user/1000/bus: No such device or address

I had to close the Terminal window to kill it.
Comment 7 HomeBoy TAZ 2022-02-13 16:24:31 CET
Hi,
I have made some tests on my side after getting into the same issue and according to them, webkit2 is not the culprit, but those two packages in their 2.34.5 version:
 * lib64webkit2gtk4.0_37
 * lib64webkit2gtk-gir4.0
I was able to downgrade them again in their 2.34.4 version and Evolution came back.
At this point, I have no more clue to help, but I can run whatever command you need for debugging.
Thanks

CC: (none) => nicolas.bachelet1

Comment 8 aguador 2022-02-13 18:53:33 CET
Glad others have contributed. Since evolution-3-38.3-3 did not work, I did a cursory check of the evo ml archive, but nothing popped out.

(Aside, was evo rebuilt before or after the webkit2 upgrade?)

This is not the first time that webkit2 changes have caused problems for evo, and probably not the last. For QA purposes it might be a good idea to make sure that webkit2 upgrades are tested against Evo in the stable release.

This will be obvious to some, but my main machine is Cauldron running evolution-3.42.4 and webkit2-2.34.5 with no problem. So whatever changes in webkit2 libraries are affecting evo, they have been solved in later evolution releases.
Comment 9 Dave Hodgins 2022-02-13 20:50:54 CET
Before installing the updates testing version ...
evolution working on my system. Help/About shows 3.38.3 (3.38.3-2.mga8).
$ rpm -qa --last|grep evolution
evolution-3.38.3-2.mga8.x86_64                2021-07-14T13:53:11 EDT
evolution-data-server-tests-3.38.3-2.mga8.x86_64 2021-07-13T18:21:06 EDT
evolution-data-server-3.38.3-2.mga8.x86_64    2021-07-13T18:21:06 EDT

After installing the updates testing version
$ rpm -qa --last|grep evolution
evolution-3.38.3-3.mga8.x86_64                2022-02-13T14:48:24 EST
evolution-data-server-tests-3.38.3-2.mga8.x86_64 2021-07-13T18:21:06 EDT
evolution-data-server-3.38.3-2.mga8.x86_64    2021-07-13T18:21:06 EDT
Still working. Help/About shows 3.38.3 (3.38.3-3.mga8)

CC: (none) => davidwhodgins

Comment 10 Dave Hodgins 2022-02-13 20:51:48 CET
$ rpm -qa --last|grep webkit|sort
lib64kf5webkit5-5.76.0-1.mga8.x86_64          2021-07-14T13:50:48 EDT
lib64proxy-webkit-0.4.17-1.mga8.x86_64        2021-07-13T18:20:30 EDT
lib64qt5webkit5-5.212.0-1.alpha4.6.mga8.x86_64 2021-07-13T18:44:05 EDT
lib64qt5webkitwidgets5-5.212.0-1.alpha4.6.mga8.x86_64 2021-07-13T18:44:05 EDT
lib64webkit2gtk4.0_37-2.34.5-1.mga8.x86_64    2022-02-09T15:30:57 EST
lib64webkit2gtk-gir4.0-2.34.5-1.mga8.x86_64   2022-02-09T15:30:57 EST
webkit2-2.34.5-1.mga8.x86_64                  2022-02-09T15:30:57 EST
webkit2-jsc-2.34.5-1.mga8.x86_64              2022-02-09T15:30:57 EST
Thomas Andrews 2022-02-13 22:01:51 CET

CC: (none) => andrewsfarm

Comment 11 Morgan Leijström 2022-02-13 22:34:51 CET
(In reply to Dave Hodgins from comment #9)
> Still working.

Maybe it is on your system using the kf5/qt5 webkit, instead of the gtk variants?
Comment 12 Dave Hodgins 2022-02-13 23:08:10 CET
That's probable. I'm using it under kde plasma.
Comment 13 Thomas Andrews 2022-02-14 01:15:47 CET
(In reply to Morgan Leijström from comment #11)
> (In reply to Dave Hodgins from comment #9)
> > Still working.
> 
> Maybe it is on your system using the kf5/qt5 webkit, instead of the gtk
> variants?

The last two times webkit2 was tested, it was on Plasma systems. With that and the above comment in mind, I just did the same tests we used on those updates in my MGA8 Gnome Vbox guest, and webkit2 passed those. 

But Evolution still crashes, confirming the suggestion in Comment 8. It looks like future tests of webkit2 should include Evolution, and should be done in both Plasma and Gnome DEs.
Comment 14 Giuseppe Merigo 2022-02-14 08:59:22 CET
Is there an easy way to revert webkit update? It seems that Epiphany (gnome web browser) doesn't start as well:

```
$ epiphany
Oops, secure memory pool already initialized
bwrap: Can't create file at /run/user/500/bus: No such device or address
```

(I'm 500 because this is a system where 500 was the starting uid for local users)

CC: (none) => g.merigo

Comment 15 Giuseppe Merigo 2022-02-14 09:05:37 CET
I just did a temporary downgrade this way if some user needs this:

```
urpmi --downgrade lib64webkit2-devel-2.34.4-1.mga8.x86_64 lib64webkit2gtk-gir4.0-2.34.4-1.mga8.x86_64 lib64webkit2gtk4.0_37-2.34.4-1.mga8.x86_64 webkit2-jsc-2.34.4-1.mga8.x86_64 webkit2-2.34.4-1.mga8.x86_64
```
Comment 16 Morgan Leijström 2022-02-14 12:13:01 CET
Thanks, Guiseppe

Status comment: (none) => Workaround: Downgrade, see c15
Summary: Evolution window does not appear after webkit2 upgrade => Evolution & Epiphany windows does not appear after webkit2 upgrade

Comment 17 Giuseppe Merigo 2022-02-14 12:35:48 CET
To avoid accidental upgrades, you can also add these (temporary) entries to /etc/urpmi/skip.list:


```
/^lib64javascriptcore-gir4.0-2/
/^lib64javascriptcoregtk4.0_18/
/^lib64webkit2-devel/
/^lib64webkit2gtk-gir4.0/
/^lib64webkit2gtk4.0_37/
/^webkit2/
/^webkit2-jsc/
```
Comment 18 Dave Hodgins 2022-02-14 17:56:26 CET
Created attachment 13131 [details]
urpmi command to test if all packages used by epiphany are installed

I logged out of kde and into gnome. Both epiphany and evolution are working
on my system.

The attached file contains a urpmi --test command I generated by running
epiphany under strace (while in gnome), and then running a script that looks
at every file accessed and does a lookup to see which (if any) package provides
that file. In this case I've excluded font packages as they are always loaded
and I have most, if not all fonts installed.

On an affected system, please run the urpmi --test command to see which, if any
packages would be installed. This will show if there is a missing dependency.
Please report back the findings.
Comment 19 Dave Hodgins 2022-02-14 17:58:06 CET
Please ensure /etc/urpmi/skip.list does not list any packages in that list,
for this test.
Comment 20 Thomas Andrews 2022-02-14 19:24:49 CET
(In reply to Dave Hodgins from comment #18)

> 
> On an affected system, please run the urpmi --test command to see which, if
> any
> packages would be installed. This will show if there is a missing dependency.
> Please report back the findings.

In a Vbox guest, Gnome's skip.lst is empty. Not going to post the entire output, just what's pertinent:

To satisfy dependencies, the following packages are going to be installed:
(test only, installation will not be actually done)
  Package                        Version      Release       Arch    
(medium "Core Release")
  breeze-icons                   5.76.0       1.mga8        noarch  
  lib64xfce4util7                4.16.0       1.mga8        x86_64  
  lib64xfconf0_3                 4.16.0       1.mga8        x86_64  
  libxfce4util-i18n              4.16.0       1.mga8        x86_64  
  xfconf                         4.16.0       1.mga8        x86_64  
67MB of additional disk space will be used.
9.3MB of packages will be retrieved.
Comment 21 Dave Hodgins 2022-02-14 19:39:31 CET
Good. It's a small list. Please install them one at a time and see if that fixes
the issue.
Comment 22 sturmvogel 2022-02-14 20:20:50 CET
I did run Daves list in my 64bit Plasma VM. At the end all packages from the list where installed but evolution still won't show a window. Console output stays:

Oops, secure memory pool already initialized
bwrap: Can't create file at /run/user/1000/bus: No such device or address
Comment 23 Thomas Andrews 2022-02-14 20:22:01 CET
Nope. 

I did breeze-icons first, and there was no change. On a hunch, I tried xfconf next, and as I halfway expected, it drew in the rest as dependencies. But still it didn't work. 

But I'm not really surprised, as those last ones are all xfce packages. Why would they be needed in Gnome?
Comment 24 Thomas Andrews 2022-02-14 20:26:17 CET
(In reply to sturmvogel from comment #22)
> I did run Daves list in my 64bit Plasma VM. At the end all packages from the
> list where installed but evolution still won't show a window. Console output
> stays:
> 
> Oops, secure memory pool already initialized
> bwrap: Can't create file at /run/user/1000/bus: No such device or address

Same output here.

And here's something interesting... I checked, and /run/user/1000/bus exists, but it is an empty file.
Comment 25 Thomas Andrews 2022-02-14 20:30:20 CET
(In reply to Thomas Andrews from comment #24)

> 
> And here's something interesting... I checked, and /run/user/1000/bus
> exists, but it is an empty file.

Hmm. It's possible that the empty /run/user/1000/bus was created when I ran Evolution before getting the webkit2 update that triggers the problem.
Comment 26 Dave Hodgins 2022-02-14 20:41:58 CET
/run/user/$UID/bus is a socket used to communicate with dbus, not a regular file.

Please enable core updates testing and then run the urpmi test command again to
see which packages from that repo are selected, and then test each of those.

As to the xfce files being required, I doubt they are. They just happened to be
loaded by epiphany if present (I have most of the desktop environments installed
on this system.

As to bwrap, $ urpmq --whatrequires bubblewrap|sort -u
bubblewrap
buildstream
flatpak
flatpak-tests
fwupd
gnome-desktop
lib64flatpak0
lib64webkit2gtk4.0_37
libwebkit2gtk4.0_37
opam
it's not required by the desktop-minimal, so may not be installed. It is
installed on my system.
Comment 27 HomeBoy TAZ 2022-02-14 20:48:47 CET
I installed all packages from the test without success, even after restarting the system. I confirm epiphany is also affected.
I enabled the testing depot, only media/core/updates_testing/lib64krb53-1.18.3-1.2.mga8.x86_64.rpm found with test script and no change after installing it
Comment 28 HomeBoy TAZ 2022-02-14 20:58:06 CET
and none of the additional packages changed anything
Comment 29 Thomas Andrews 2022-02-14 21:50:57 CET
I used qarepo with "*.*" as the rpm list, which should have downloaded the contents of all testing repos to my local repo. Then I ran Dave's test again, and it came up with this:

installing lib64krb53-1.18.3-1.2.mga8.x86_64.rpm from //home/tom/qa-testing/x86_64
Preparing...                     #############################################
Installation is possible

Going back now to install that package...
Comment 30 Dave Hodgins 2022-02-14 21:58:19 CET
I've managed to recreate the problem on one of test installs.

In that install, both epiphany and evolution fail when run as a regular user,
but work if run as root (using "su -" and then starting the application).

So it's a permission problem somewhere, not a package installation issue.

I'm now trying to figure out what permission is different and why.
Comment 31 Thomas Andrews 2022-02-14 22:00:23 CET
Updated just that package, but still no change with running Evolution.
Comment 32 Thomas Backlund 2022-02-14 22:35:43 CET
Looks like upstream is aware of the breakages, and are queueing up fixes for a 2.34.6

https://trac.webkit.org/wiki/WebKitGTK/2.34.x#Proposedmergesfor2.34.6
Comment 33 Dave Hodgins 2022-02-14 22:48:04 CET
Can someone else confirm that it works running as root when it fails for the
regular user?
Comment 34 Dave Hodgins 2022-02-14 22:51:38 CET
Just fyi, the install I have where they are working started as a Mageia 3
install, upgraded each release since then to mga8, so it has config files
with settings from old releases.

I don't see any obvious permission issues when it fails in my other test
install running under strace.
Comment 35 sturmvogel 2022-02-14 23:00:17 CET
Evolution opens with sudo. Still throws some (different) errors:

Oops, secure memory pool already initialized
Oops, secure memory pool already initialized

(evolution:4770): GLib-GIO-WARNING **: 23:02:04.525: Your application did not unregister from D-Bus before destruction. Consider using g_application_run().
Comment 36 HomeBoy TAZ 2022-02-15 07:51:47 CET
What I can tell is that only lib64webkit2gtk4.0_37 and lib64webkit2gtk-gir4.0 generate this issue, not the webkit2-2.34.5-1.mga8 package because I was able to update it without problem.

Those 2 packages do not have much files, maybe easier to debug?

On my machine, evolution does not start with root, it says it is already running using another user account. I checked and did not saw any running process. I do not know if it is related or a different issue, I never had such idea to run it as root ;)
Comment 37 Giuseppe Merigo 2022-02-15 08:18:26 CET
(In reply to Dave Hodgins from comment #18)
> Created attachment 13131 [details]
> urpmi command to test if all packages used by epiphany are installed
> 

On my test vm (which shows the same behavior of my notebook):

  Package                        Version      Release       Arch    
(medium "Core Release")
  GConf2                         3.2.6        19.mga8       x86_64  
  enchant2                       2.2.15       1.mga8        x86_64  
  lib64GConf2_4                  3.2.6        19.mga8       x86_64 

Still both epiphany and evolution do not start.

Note that if you start epiphany from console and wait long enough, this message gets printed on stdout:

```
Failed to register: Timeout was reached
```

while evolution does not seem to do that (maybe I should wait longer than a few minutes)?

I can confirm that both evolution and epiphany run as root, maybe there are some permission issues on the dbus?
Comment 38 Giuseppe Merigo 2022-02-15 08:32:41 CET
I did some research on the bwrap error, it seems that it's related to sandboxing as the sandbox is unable to escape to other paths (see this for an example https://github.com/flatpak/flatpak/issues/3306). 

But... I do not think evolution and epiphany rpms are sandboxed by default, maybe the new libwebkit version do its own sandboxing without the software depending on it knowledge?

Maybe this sandboxing does not apply to anything but only to dbus? And this has something to do with the fact that by using root everything works?

In this case we shouldn't be the only ones with this issue.
Comment 39 HomeBoy TAZ 2022-02-15 20:28:41 CET
Regarding Giuseppe's previous note, I waited a longer time but no other error shown, tried epiphany and evolution
Comment 40 Ricard Alfe 2022-02-16 09:57:05 CET
Sorry for my english, translation made by google.

I too have this problem and have temporarily fixed it by running the command in Konsole :
su - "my username"
then you run "evolution" and it's working fine.
    I've been with this error for a while so I'm going to reinstall. It has to be a permissions error.

CC: (none) => ricardalfe

Comment 41 Giuseppe Merigo 2022-02-16 10:11:08 CET
(In reply to Ricard Alfe from comment #40)
> I too have this problem and have temporarily fixed it by running the command
> in Konsole :
> su - "my username"
> then you run "evolution" and it's working fine.
>     I've been with this error for a while so I'm going to reinstall. It has
> to be a permissions error.

Yes, if you run it via root it works fine.
Don't reinstall anything since I've made tests with brand new users, and the problem is the same, if you do the webkit upgrade, at least epiphany (Gnome Web) and Evolution stop working.
Comment 42 Giuseppe Merigo 2022-02-16 11:50:45 CET
I did some more research about the bwrap command, I assumed it was part of webkit, but it doesn't:

```
$ rpm -qf /usr/bin/bwrap
bubblewrap-0.4.1-1.mga8
$ rpm -qi bubblewrap
Name        : bubblewrap
Version     : 0.4.1
Release     : 1.mga8
Architecture: x86_64
Install Date: Fri 09 Apr 2021 12:40:19 PM CEST
Group       : Security
Size        : 87790
License     : LGPLv2+
Signature   : RSA/SHA256, Mon 30 Mar 2020 05:20:12 PM CEST, Key ID b742fa8b80420f66
Source RPM  : bubblewrap-0.4.1-1.mga8.src.rpm
Build Date  : Mon 30 Mar 2020 05:19:33 PM CEST
Build Host  : localhost
Packager    : daviddavid <daviddavid>
Vendor      : Mageia.Org
URL         : https://github.com/projectatomic/bubblewrap
Summary     : Core execution tool for unprivileged containers
Description :
Bubblewrap (/usr/bin/bwrap) is a core execution engine for unprivileged
containers that works as a setuid binary on kernels without
user namespaces.
```

So maybe after the upgrade on webkit this software is not working anymore and needs an update or a configuration upgrade?
Comment 43 Ricard Alfe 2022-02-16 12:07:03 CET
(In reply to Giuseppe Merigo from comment #41)
> (In reply to Ricard Alfe from comment #40)
> > I too have this problem and have temporarily fixed it by running the command
> > in Konsole :
> > su - "my username"
> > then you run "evolution" and it's working fine.
> >     I've been with this error for a while so I'm going to reinstall. It has
> > to be a permissions error.
> 
> Yes, if you run it via root it works fine.
> Don't reinstall anything since I've made tests with brand new users, and the
> problem is the same, if you do the webkit upgrade, at least epiphany (Gnome
> Web) and Evolution stop working.

you are not working as root, you are working as your user. If you make an "id" you will see that you are not "root id:0" but "id:1000" or a higher id
Comment 44 Ricard Alfe 2022-02-16 12:15:21 CET

At the moment I have already reinstalled, I have started with a clean user, I have removed nvidia proprietary drivers. All in vain, no positive results
Comment 45 Morgan Leijström 2022-02-16 12:40:30 CET
What if you run it under Plasma, using kf5/qt5 webkit?
comment 11 - 13
Gerhard Proess 2022-02-16 13:37:03 CET

CC: (none) => gp

Comment 46 Ricard Alfe 2022-02-16 13:42:34 CET
I work at
Sistema operativo: Mageia 8
Versión de KDE Plasma: 5.20.4
Versión de KDE Frameworks: 5.76.0
Versión de Qt: 5.15.2
Versión del kernel: 5.15.23-desktop-1.mga8
Tipo de OS: 64 bits
Procesadores: 16 × Intel® Core™ i7-10700K CPU @ 3.80GHz
Procesador gráfico: NVIDIA GeForce GTX 960/PCIe/SSE2
webkit2 2.34.5


I have also installed gnome and it doesn't work either
Comment 47 Ricard Alfe 2022-02-16 13:45:15 CET
the only thing I have achieved is that it asks me for the password of the box when installing policycoreutils-sandbox
Comment 48 Ricard Alfe 2022-02-16 14:16:54 CET
I have found a solution by visiting these pages:

https://github.com/netblue30/firejail/issues/3647
https://bbs.archlinux.org/viewtopic.php?id=259928

just have to run 

WEBKIT_FORCE_SANDBOX=0 firejail evolution

It has to do with sandbox and bwrap
Comment 49 Giuseppe Merigo 2022-02-16 14:30:40 CET
I do not have firejail but I can confirm that starting from command line both epiphany and evolution like this works for me in my test system. At least the applications start.

```
$ WEBKIT_FORCE_SANDBOX=0 epiphany
```

```
$ WEBKIT_FORCE_SANDBOX=0 evolution
```

I think that this is disabling some security feature so it's another workaround, maybe less severe that not updating the files.

I'll try the workaround in my main machine as soon as possible to see if there are any issues with a real life scenario.
Comment 50 Ricard Alfe 2022-02-16 14:43:42 CET
I do not have firejail but I can confirm that starting from command line both epiphany and evolution like this works for me in my test system. At least the applications start.

```
$ WEBKIT_FORCE_SANDBOX=0 epiphany
```

```
$ WEBKIT_FORCE_SANDBOX=0 evolution
```

I think that this is disabling some security feature so it's another workaround, maybe less severe that not updating the files.

I'll try the workaround in my main machine as soon as possible to see if there are any issues with a real life scenario.
Comment 51 Ricard Alfe 2022-02-16 14:46:44 CET
(In reply to Ricard Alfe from comment #50)
> I do not have firejail but I can confirm that starting from command line
> both epiphany and evolution like this works for me in my test system. At
> least the applications start.
> 
> ```
> $ WEBKIT_FORCE_SANDBOX=0 epiphany
> ```
> 
> ```
> $ WEBKIT_FORCE_SANDBOX=0 evolution
> ```
> 
> I think that this is disabling some security feature so it's another
> workaround, maybe less severe that not updating the files.
> 
> I'll try the workaround in my main machine as soon as possible to see if
> there are any issues with a real life scenario.



yes, it has to do with setgui permissions, I think, it prevents programs from having scalable permissions to root. Creating small cages for programs with extra permissions.
Comment 52 Dave Hodgins 2022-02-16 18:29:48 CET
(In reply to Ricard Alfe from comment #40)
> Sorry for my english, translation made by google.
> 
> I too have this problem and have temporarily fixed it by running the command
> in Konsole :
> su - "my username"
> then you run "evolution" and it's working fine.
>     I've been with this error for a while so I'm going to reinstall. It has
> to be a permissions error.

Confirmed. On my test install where it's failing
[tester@x8v ~]$ epiphany 
Oops, secure memory pool already initialized
bwrap: Can't create file at /run/user/1001/bus: No such device or address
Killed (using killall -9 epiphany on another konsole).

[tester@x8v ~]$ su -l tester
Password: 
[tester@x8v ~]$ epiphany (It Works!)
Oops, secure memory pool already initialized
Oops, secure memory pool already initialized

This also helps explain why it's working in my main install, where I use
run level 3 and then startx.

WEBKIT_FORCE_SANDBOX is not set in either the working or failing environments.

Is anyone having it fail using a dm other then sddm?

Before using the "su -l tester", the following environment variables related
to dbus are set ...
$ grep -i bus ??su.txt
b4su.txt:KONSOLE_DBUS_SERVICE=:1.75
b4su.txt:KONSOLE_DBUS_SESSION=/Sessions/1
b4su.txt:DBUS_SESSION_BUS_ADDRESS=unix:path=/run/user/1001/bus
b4su.txt:KONSOLE_DBUS_WINDOW=/Windows/1

After running "su -l tester", none of those are set.

Based on that, I tried "unset DBUS_SESSION_BUS_ADDRESS" and that is all that
is needed to get it working.
Comment 53 Dave Hodgins 2022-02-16 18:48:32 CET
Using run level 3 and then startx in the test install, DBUS_SESSION_BUS_ADDRESS
is set ..
DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-w2jzheWwwu,guid=ca072b7fe92a48bf4e812a84620d3717

The /tmp directory does not contain any dbus file or socket.
Comment 54 Dave Hodgins 2022-02-16 18:55:38 CET
Using run level 3 and startx in my test install with
DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-w2jzheWwwu,guid=ca072b7fe92a48bf4e812a84620d3717
epiphany works.

Using run level 5 in my test install with
DBUS_SESSION_BUS_ADDRESS=unix:path=/run/user/1001/bus
epiphany fails.
Comment 55 Dave Hodgins 2022-02-16 19:27:43 CET
So the question becomes why is DBUS_SESSION_BUS_ADDRESS different when using
run level 5 versus run level 3 and startx, how does webkit2 fit into it, and
how to fix it properly?
Comment 56 Ricard Alfe 2022-02-16 19:42:11 CET
I don't quite understand why you use runlevel and not systemctl start multi-user.target . I was resisting using systemctl until I realized that some configurations did not start
Comment 57 Thomas Backlund 2022-02-16 20:33:00 CET
(In reply to Dave Hodgins from comment #55)
> So the question becomes why is DBUS_SESSION_BUS_ADDRESS different when using
> run level 5 versus run level 3 and startx, how does webkit2 fit into it, and
> how to fix it properly?

the bug is already identified upstream and a fix is already merged in trunk and assigned for upcoming 2.34.6..
If they dont release it soon-ish I'll probably look into backporting it ...
Comment 58 Thomas Andrews 2022-02-16 20:58:18 CET
(In reply to Dave Hodgins from comment #52)

> 
> Is anyone having it fail using a dm other then sddm?
> 
Sorry I wasn't quicker with this. Probably don't need this question answered any more, but...

My Gnome Vbox guest is using gdm.
Comment 59 aguador 2022-02-16 21:23:01 CET
And, FWIW, my Enlightenment install uses LXDE.
Comment 60 Dave Hodgins 2022-02-16 21:27:19 CET
(In reply to Ricard Alfe from comment #56)
> I don't quite understand why you use runlevel and not systemctl start
> multi-user.target . I was resisting using systemctl until I realized that
> some configurations did not start

It's easier remember. I just edit the kernel parameters and add " 3", instead
of having to remember the systemd way to do the same thing.

systemd translates a " 1", " 3", or " 5" to the target automatically.
Comment 61 Ricard Alfe 2022-02-17 00:15:14 CET
I understand you for me it is also easier
Comment 62 Giuseppe Merigo 2022-02-17 08:12:58 CET
(In reply to Dave Hodgins from comment #55)
> So the question becomes why is DBUS_SESSION_BUS_ADDRESS different when using
> run level 5 versus run level 3 and startx, how does webkit2 fit into it, and
> how to fix it properly?

I think that if you start X on your own you don't have a running dbus, so the application starts a local dbus by its own. This would answer the question about why it's in a different location and why it works because it's probably started in the same sandbox as the application itself.

This of course cannot be a solution as people not using X11 (like wayland) would still not be able to use Evolution or Epiphany.
Comment 63 Giuseppe Merigo 2022-02-17 08:15:15 CET
(In reply to Thomas Backlund from comment #57)

> the bug is already identified upstream and a fix is already merged in trunk
> and assigned for upcoming 2.34.6..
> If they dont release it soon-ish I'll probably look into backporting it ...

So webkit developers actually chose to break end user applications with a security update?
Comment 64 aguador 2022-02-17 08:59:43 CET
Thomas has the most sensible answer as it seems that a fix will be available soonish and downgrading does take care of the problem in the interim.

However, given the extensive back and forth and interest in understanding this, I will throw out, for those interested in pursuing it, something that this discussion suggests might be related. At some point recently, before the last webkit update, Evolution messages in the outbox stopped being visible or accessible in the message pane, even though the counter showed their presence. As with this bug, Cauldron has continued to work properly with the latest packages.
Comment 65 William Shields 2022-02-17 11:00:51 CET
If a solution might take a while then another possible workaround that I have found might be of interest. It involves editing the 'Exec' statement in the .desktop files of epiphany (org.gnome.epiphany.desktop) and evolution (org.gnome.evolution.desktop) (both in /usr/share/applications) to add WEBKIT_FORCE_SANDOX=0 before calling the application e.g.

Exec=env WEBKIT_FORCE_SANDBOX=0 evolution %U

This works on my laptop with a standard Gnome desktop install from live media. It has the advantage that I (probably) wont have to remember to undo the changes in order to get the fix to work when it is released.

As a regular user of evolution and sometime user of epiphany I am very grateful for the time and effort that the contributors have put into solving this problem. When I saw the reference to Sand Boxes I thought perhaps it might just have been a problem with the build of one or other of the libraries failing to switch this off. Apparently the call to bwrap comes from webkit2gtk code so my guess would have been one of the lib64webkit2gtk* mentioned in comment 7

CC: (none) => william.shields

Comment 66 Giuseppe Merigo 2022-02-17 14:21:40 CET
(In reply to William Shields from comment #65)
> If a solution might take a while then another possible workaround that I
> have found might be of interest. It involves editing the 'Exec' statement in
> the .desktop files of epiphany (org.gnome.epiphany.desktop) and evolution
> (org.gnome.evolution.desktop) (both in /usr/share/applications) to add
> WEBKIT_FORCE_SANDOX=0 before calling the application e.g.
> 
> Exec=env WEBKIT_FORCE_SANDBOX=0 evolution %U
> 

This is what I'm doing but it's not a workaround for everyone. Many desktop systems, such as XFce, have custom desktop files to benefit from custom shortcuts (send mail, save draft, check for new mail and so on).
For instance in my test VM I use xfce and I don't even know how to insert an environment variable there as the desktop file for evolution doesn't seem to be standard.
Comment 67 Thomas Backlund 2022-02-17 21:43:06 CET
 webkit2 2.34.6 is now released upstream and is submitted to updates_testing
Thomas Backlund 2022-02-17 21:50:40 CET

Depends on: (none) => 30064

Comment 68 Thomas Backlund 2022-02-18 07:24:05 CET
Can the affected users test if the 2.34.6 update actually resolves the issue, and report back here or in bug 30064
Comment 69 sturmvogel 2022-02-18 07:42:31 CET
Updated the three webkit packages from updates_testing.

Plasma x86_64

Epiphany opens correctly (opened some sites and surfed a little bit, all ok)
Evolution opens correctlty (did some settings and test setups, all ok)


MGA8 64bit OK
Comment 70 sturmvogel 2022-02-18 07:45:29 CET
To be more precise, installed
lib64webkit2gtk4.0_37-2.34.6-1.mga8.x86_64.rpm
lib64webkit2gtk-gir4.0-2.34.6-1.mga8.x86_64.rpm
webkit2-2.34.6-1.mga8.x86_64.rpm
Comment 71 HomeBoy TAZ 2022-02-18 08:03:29 CET
Evolution and Epiphany both start using 2.34.6 on my computer, with and without a restart. Sounds fixed in my VM environment
Comment 72 Giuseppe Merigo 2022-02-18 08:18:30 CET
I can confirm the new version works on my test VM mga8 64 bits, under XFCE. 
I'm trying these new packages in a bit on my working notebook (also mga8, 64 bits but under Gnome)
Comment 73 Giuseppe Merigo 2022-02-18 08:32:17 CET
I confirm that with packages in comment #70 and their dependencies, both epiphany and evolution work again without any other modification (such as environment entries), on a mga8 x86_64 under Gnome Shell.
Comment 74 Thomas Backlund 2022-02-18 10:33:29 CET
Awesome.

Thanks for confirming.

I'll flush out the update then
Comment 75 Ricard Alfe 2022-02-18 12:33:52 CET
all good in the update, thank you very much
Comment 76 Thomas Backlund 2022-02-18 16:18:44 CET
An update for this issue has been pushed to the Mageia Updates repository.

https://advisories.mageia.org/MGASA-2022-0075.html

Resolution: (none) => FIXED
Status: NEW => RESOLVED


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