Closed Bug 2081619 Opened 1 month ago Closed 1 month ago

Firefox Nightly doesn't open in non kiosk mode

Categories

(Core :: Widget: Gtk, defect)

Firefox 160
defect

Tracking

()

RESOLVED DUPLICATE of bug 2081642

People

(Reporter: roland1705, Unassigned)

References

Details

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:160.0) Gecko/20100101 Firefox/160.0

Steps to reproduce

Since the rpm version 0:160.0a1-20261012185752 of Firefox Nightly, the main window of Firefox never shows up on Fedora Workstation 44 (KDE Plasma).

Actual results

In the task bar, the icon shows up but no window is available. Enabling kiosk mode from the terminal does show the window in full screen without the tab bar. Safe mode didn't fix the issue.

Expected results

The window should have appeared just like it does on 0:160.0a1-20261012091945 downwards.

I also tried with MOZ_DISABLE_WAYLAND=1 and MOZ_ENABLE_WAYLAND=1 without success.

The Bugbug bot thinks this bug should belong to the 'Core::Widget: Gtk' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Component: Untriaged → Widget: Gtk
Product: Firefox → Core

This looks like a regression from bug 2080307.

The reporter's range (Nightly 20261012091945 good, 20261012185752 bad) covers 51 commits. The only Wayland change among them is bug 2080307, "[Wayland] Apply toplevel opaque region changes immediately". The only other widget/gtk change is bug 2080242, which just adds an #include.

Bug 2081642 looks like the same issue, from a different reporter on Fedora / KDE Plasma / Wayland with Nightly 160. There the window appears with MOZ_ENABLE_WAYLAND=0, so this only happens on Wayland. (In comment 1, MOZ_DISABLE_WAYLAND=1 probably had no effect because the invisible instance was still running, so the new launch was passed to the existing process.)

Possible mechanism, from reading the code (not verified on KWin): bug 2080307 makes nsWindow::UpdateOpaqueRegionInternal() call the new nsWindowWayland::SetToplevelOpaqueRegion(). That sets the opaque region on the toplevel wl_surface and then calls ForceToplevelCommit(). UpdateOpaqueRegionInternal() also runs from toplevel_window_size_allocate_cb, which fires before the toplevel is mapped. So the first call now creates the commit sub-surface (mCommitWindow) and forces a toplevel commit before the window is mapped, outside GTK's normal sequence. KWin then appears to never map the window. Kiosk mode, which starts fullscreen, still shows it.

Skipping the direct commit in SetToplevelOpaqueRegion() until the toplevel is mapped might be enough.

Setting Bug 2080307 as see also until it's confirmed as the regressor.

:stransky, could you take a look? I would like to back out Bug 2080307, but I need some confirmation first.

Flags: needinfo?(stransky)
See Also: → 2081642, 2080307

(In reply to Roland Greffe from comment #1)

I also tried with MOZ_DISABLE_WAYLAND=1 and MOZ_ENABLE_WAYLAND=1 without success.

The right thing to try would be MOZ_ENABLE_WAYLAND=0

I guess bug 2080307 is a bit suspicious indeed, maybe forcing the commit too early breaks GTK's clock? That's happened in the past I think...

Severity: -- → S3
Severity: S3 → S2

(I couldn't repro this off-hand, but on Plasma 6.7.91

Status: UNCONFIRMED → RESOLVED
Closed: 1 month ago
Duplicate of bug: 2081642
Resolution: --- → DUPLICATE
Flags: needinfo?(stransky)
You need to log in before you can comment on or make changes to this bug.