Thanks, I’ve now reported this at: https://gitlab.gnome.org/GNOME/gnome-shell/issues/287

Background

I’m using Ubuntu 18.04 with Gnome Shell 3.28.1, and was forwarded here from Launchpad, see bug #1767806 there.

The issue

When another window is maximized or tiled to the left side of the screen, re-opening an application whose window was previously tiled to the left side does not restore that windows’ previous position. Instead such windows open at varying distances from the dock, untiled.

Expected behavior

Steps to reproduce (with a newly created user)

1. open Firefox, drag the window to the left edge of the screen to tile it there (filling up the left half of horizontal screen space)
2. close Firefox
3. open Nautilus, make sure that the window floats (i.e. is unmaximized and not tiled to either side of screen)
4. open Firefox

What happens

As expected, Firefox’s window is tiled to the left edge of screen, immediately to the right side of the dock: screenshot.

Unexpected behavior

Steps to reproduce (with a newly created user)

1. open Firefox, drag the window to the left edge of the screen to tile it there (filling up the left half of horizontal screen space)
2. close Firefox
3. open Nautilus, maximize its window
4. open Firefox

What happens

Firefox’s window opens slightly to the right off the right edge of the dock (with a gap between the dock and the window): screenshot.

What I expect to happen

I expect Firefox’s window to be tiled to the left edge of screen, immediately to the right side of the dock, just as it did following the first recipe above.

Further information

• Either maximizing or tiling Nautilus’ window to the left edge of the screen (at step 3) triggers the issue. Tiling it to the right edge of the screen does not.
• Using Gnome Terminal instead of Nautilus triggers the issue just as well and I suspect any other application will, though I’ve only systematically tested these two so far.
• From battling with this in my daily use I also know that the distance of incorrectly restored windows from the dock varies: different applications open at different distances. The precise mechanics of this still elude me, though the distances do appear to be multiples of 25 pixels.

When another window is maximized or tiled to the left side of the screen, re-opening an application whose window was previously tiled to the left side does not restore that windows’ previous position. Instead such windows open at varying distances from the dock, untiled.

= Steps to see expected behavior (with a newly created user) =

1. open Firefox, drag the window to the left edge of the screen to tile it there (filling up the left half of horizontal screen space)
2. close Firefox
3. open Nautilus, make sure that the window floats (i.e. is unmaximized and not tiled to either side of screen)
4. open Firefox

== What happens ==
As expected, Firefox’s window is tiled to the left edge of screen, immediately to the right side of the dock.

= Steps to reproduce unexpected behavior (with a newly created user) =
1. open Firefox, drag the window to the left edge of the screen to tile it there (filling up the left half of horizontal screen space)
2. close Firefox
3. open Nautilus, *maximize its window*
4. open Firefox

== What happens ==
Firefox’s window opens slightly to the right off the right edge of the dock.

== What I expect to happen ==
I expect Firefox’s window to be tiled to the left edge of screen, immediately to the right from the dock, just as it did following the first recipe above.

= Further information =
Either maximizing or tiling Nautilus’ window to the left edge of the screen (at step #3) triggers the issue. Tiling it to the *right* edge of the screen does not.

Using Gnome Terminal instead of Nautilus triggers the issue just as well, I suspect any other application will though I’ve only systematically tested these two so far.

From battling with this in my daily use I also know that the distance of incorrectly restored windows from the dock varies: different applications open at different distances. The precise mechanics of this still elude me, though the distances do appear to be multiples of 25 pixels.

If the services run under the ’git’ user, why is it that adding the ’git’ user to the ’redis’ group (with access to the socket) isn’t enough to give it access to the socket? Sorry if this is a dumb question.

Käytän Polar Flow’ta treenieni kirjaamiseen, ja ne tuodaan sieltä Heiaheia-tililleni automaattisesti. Tämä toimii muutoin ihan hyvin, mutta sisäpyöräily kirjautuu jostain syystä Heiaheiassa tavalliseksi pyöräilyksi. Olen ruukannut korjata nuo Heiaheian puolella jälkikäteen, mutta olisi kätevää jos tämäkin toimisi automaattisesti niin kuin pitää.

Liitteenä on esimerkkikuva tämänpäiväisestä treenistä Flow’ssa ja Heiaheiassa.

(En ole varma onko vika Heiaheian vai Polarin päässä, mutta ajattelin kokeilla ensin tätä kautta.)

Kuvakaappaus-HH-vs-Polar.png

Red Hat issue: https://bugzilla.redhat.com/show_bug.cgi?id=1484411 ”Running wget with -O and -q in the background yields a file wget-log”

Related but (FWICT) not precisely the same issues upstream:

* https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=888691 ”wget -b produces empty wget-log file”
* https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=874590 ”wget: creates log file when run in the background”, forwarded to https://savannah.gnu.org/bugs/?51181 ”Unexpected ”Redirecting output to ’wget-log’.””

== Steps to reproduce ==
`wget -q https://www.google.com/ &`

== What happens ==
wget says it’s ”Redirecting output to ’wget-log'” and creates an empty wget-log file in the current directory.

== What I expect to happen ==
No wget-log file to be created.

== Details ==
Wget has a -b (–background) option, for which the man page says: ”If no output file is specified via the -o, output is redirected to wget-log.” There’s also the -q (–quiet) option, for which the man page says: ”Turn off Wget’s output.”

When both are given, wget goes to background, and no wget-log gets created, as expected.

When only -b is given, wget goes to background, and creates wget-log, also as expected.

When only -q is given, wget runs in the foreground, and no wget-log is created, still as expected.

When only -q is given, but & is added to the end of the command line, wget goes to background, but also creates wget-log, which I find unexpected (and inconsistent).

I’ve had this happen a couple of times now since upgrading to 18.04: soon (perhaps within a minute) after logging in, Gnome session ends abruptly and I’m thrown back to the login screen. IIRC, in both instances this occurred on the first login after boot, so it does not reoccur on the subsequent re-login, and not on every login (very rarely in fact), and for now I have no better steps to reproduce this other than ”boot, log in”.

Bug #1663839 (reported against 17.04) seems similar, as does bug #1731428 (reported against 17.10).

journalctl output during the issue:

Apr 16 07:00:30 saegusa gnome-session[4342]: gnome-session-binary[4342]: WARNING: Application ’org.gnome.SettingsDaemon.Color.desktop’ failed to register before timeout
Apr 16 07:00:30 saegusa gnome-session-binary[4342]: WARNING: Application ’org.gnome.SettingsDaemon.Color.desktop’ failed to register before timeout
Apr 16 07:00:30 saegusa gnome-session-binary[4342]: Unrecoverable failure in required component org.gnome.SettingsDaemon.Color.desktop
Apr 16 07:00:30 saegusa gnome-session[4342]: gnome-session-binary[4342]: CRITICAL: We failed, but the fail whale is dead. Sorry….
Apr 16 07:00:30 saegusa gnome-session-binary[4342]: CRITICAL: We failed, but the fail whale is dead. Sorry….

In addition to Totem, Firefox also exhibits this with fullscreen HTML5 video. Here’s a good test: https://www.youtube.com/watch?v=5xkNy9gfKOg

My bug #1762400 got marked as a duplicate for this one, but I don’t have Dropbox installed.