Seems I've gotten pretty good at solving these Chrome UI bugs. I'll copy my fix from the Chromium bug tracker to help my fellow Arch users out.
OK I've found the root of this bug and it's not really Chrome's fault. There's a dirty hack in the Adwaita theme to work around a bug in an old version of LibreOffice. To fix, look in /usr/share/themes/Adwaita/gtk-2.0/gtkrc and look for this section of code:
style "menu_item"
{
xthickness = 2
ythickness = 4
# HACK: Gtk doesn't actually read this value
# while rendering the menu items, but Libreoffice
# does; setting this value equal to the one in
# fg[PRELIGHT] ensures a code path in the LO theming code
# that falls back to a dark text color for menu item text
# highlight. The price to pay is black text on menus as well,
# but at least it's readable.
# See https://bugs.freedesktop.org/show_bug.cgi?id=38038
bg[SELECTED] = @selected_fg_color
bg[PRELIGHT] = @selected_bg_color
fg[PRELIGHT] = @selected_fg_color
text[PRELIGHT] = @selected_fg_color
Change bg[SELECTED] = @selected_fg_color to @selected_bg_color and the bug should be fixed for Chrome. Not sure if this breaks recent versions of LibreOffice or not.
The reason this only affects Chrome is that every other GTK application (with the exception of LibreOffice apparently) uses the bg[PRELIGHT] field while Chrome uses the bg[SELECTED] field.
The reason this broke with version 36 is there was also a section in that same file that reverted this behavior specifically for Chrome.
style "chrome_menu_item"
{
bg[SELECTED] = @selected_bg_color
}
And further down where this is applied:
# Chrome/Chromium
class "ChromeGtkFrame" style "chrome-gtk-frame"
widget_class "*Chrom*Button*" style "button"
widget_class "*<GtkCustomMenu>*<GtkCustomMenuItem>*" style "chrome_menu_item"
# We use this weird selector to target an offscreen entry as created
# by Chrome/Chromium to derive the style for its toolbar
widget_class "<GtkEntry>" style "standalone_entry"
I'm guessing Google changed the name of its frame from ChromeGtkFrame (or maybe they changed the name of their menu.) In any case, some change to Chrome broke this reverse-dirty-hack and caused the bug.
Search Criteria
Package Details: google-chrome 131.0.6778.85-1
Package Actions
Git Clone URL: | https://aur.archlinux.org/google-chrome.git (read-only, click to copy) |
---|---|
Package Base: | google-chrome |
Description: | The popular web browser by Google (Stable Channel) |
Upstream URL: | https://www.google.com/chrome |
Keywords: | chromium |
Licenses: | custom:chrome |
Submitter: | None |
Maintainer: | gromit |
Last Packager: | gromit |
Votes: | 2248 |
Popularity: | 10.09 |
First Submitted: | 2010-05-25 20:25 (UTC) |
Last Updated: | 2024-11-19 19:19 (UTC) |
Dependencies (12)
- alsa-lib
- gtk3 (gtk3-no_deadkeys_underlineAUR, gtk3-classicAUR, gtk3-classic-xfceAUR, gtk3-patched-filechooser-icon-viewAUR)
- libcups (libcups-gitAUR, cups-gitAUR, libcups-gssapiAUR)
- libxss
- libxtst
- nss (nss-hgAUR)
- ttf-liberation (ttf-defenestrationAUR)
- xdg-utils (busking-gitAUR, xdg-utils-slockAUR, mimiAUR, mimi-gitAUR, xdg-utils-handlrAUR, openerAUR, xdg-utils-mimeoAUR, mimejs-gitAUR)
- gnome-keyring (gnome-keyring-gitAUR) (optional) – for storing passwords in GNOME keyring
- kdialog (kdialog-gitAUR) (optional) – for file dialogs in KDE
- kwallet (kwallet-gitAUR) (optional) – for storing passwords in KWallet
- pipewire (pipewire-gitAUR, pipewire-full-gitAUR) (optional) – WebRTC desktop sharing under Wayland
Required by (40)
- bitwarden-chromium (optional)
- captive-browser-git (optional)
- chrome-extension-bitwarden-git (optional)
- chrome-extension-ocrs-git
- chromedriver (optional)
- chromium-extension-adnauseam (optional)
- chromium-extension-autoscroll (optional)
- chromium-extension-plasma-integration (optional)
- chromium-extension-runet-censorship-bypass (optional)
- chromium-material-icons-for-github-bin (optional)
- chromium-vencord (optional)
- chromium-vencord-bin (optional)
- chromium-vencord-git (optional)
- dedao-dl-bin (optional)
- endpoint-verification-chrome
- endpoint-verification-minimal
- ff2mpv-go-git (optional)
- ff2mpv-rust (optional)
- hub-kids (optional)
- hub-young (optional)
- ice-ssb (optional)
- ice-ssb-git (optional)
- kget-integrator-chrome (optional)
- lastpass (optional)
- marp-cli (optional)
- nfauthenticationkey (optional)
- pearson-reader-plus-full-lang (optional)
- pennywise-bin (optional)
- pt-plugin-plus-bin (optional)
- pt-plugin-plus-git (optional)
- python-nativemessaging-ng (optional)
- python-webdriver-manager (check)
- quick-n-easy-web-builder-10 (optional)
- sshcode-bin (optional)
- uget-integrator-chrome (optional)
- upload-gphotos (optional)
- web-media-controller-mpris (optional)
- web-media-controller-mpris-git (optional)
- webchanges (optional)
- webui-aria2-git (optional)
Sources (3)
Latest Comments
« First ‹ Previous 1 .. 96 97 98 99 100 101 102 103 104 105 106 .. 157 Next › Last »
weirddan455 commented on 2014-07-20 22:02 (UTC)
<deleted-account> commented on 2014-07-18 17:46 (UTC)
This bug now also affects the Stable channel:
https://code.google.com/p/chromium/issues/detail?id=378090
Det commented on 2014-07-10 19:04 (UTC)
No I mean as in it's not necessary to put in "Don't believe me?" in the midst of all that, when you clearly know what you're talking about.
weirddan455 commented on 2014-07-10 19:02 (UTC)
...What don't you believe? Did setting XCURSOR_THEME not fix Chrome's cursor for you?
Det commented on 2014-07-10 18:54 (UTC)
I _don't_ believe you.
weirddan455 commented on 2014-07-10 16:01 (UTC)
On the mouse cursor issue, the reason no one is fixing it is because it's not a Chrome bug. You should be filing a bug report for your DE for not correctly setting the XCURSOR_THEME environment variable (and optionally XCURSOR_SIZE if you need a custom size) That's the standard way for every X application to know which cursor to use. On KDE, this variable is set in the /usr/bin/startkde script. Don't believe me? Try launching any X application from the console, setting that variable to a different font. For example "XCURSOR_THEME=Oxygen_Blue firefox" Firefox used to use Oxygen_White like the rest of the system but now it uses Oxygen_Blue. Chrome works the same way. The old version may have grabbed your cursor theme setting from the GTK theme but since the new version no longer uses GTK it's not going to look there.
On the whole screen going blank, I would try disabling GPU acceleration. In Settings > Advanced, uncheck "Use hardware acceleration when available." You can also disable Flash's GPU acceleration by right clicking a Flash widget and clicking settings. You could also just disable Flash in about:plugins to rule it out entirely (may want to try this first.)
catalin.hritcu commented on 2014-06-26 08:19 (UTC)
Managed to find something in dmesg that seems correlated with the failures:
https://gist.github.com/catalin-hritcu/bd0a5b374a528b382cad
My video card is a Intel HD Graphics 4600, on a Lenovo T440p. There is also an NVidia card inside but it's disabled.
catalin.hritcu commented on 2014-06-26 08:01 (UTC)
I'm experiencing a new problem with the latest version: every now and then the whole screen starts going blank for a while. Not just the Chrome window but the whole screen. And it doesn't even matter whether the Chrome window is active or just somewhere in the background. All I could find to stop the problem is to restart Chrome, but eventually it comes back. I guess it's related to the new Aura thing, but I didn't manage to get any output in the console that would give me more information to be able to file a good bug report. I'm using MATE.
Det commented on 2014-06-23 19:47 (UTC)
Done.
Pinned Comments
gromit commented on 2023-04-15 08:22 (UTC) (edited on 2023-05-08 21:42 (UTC) by gromit)
When reporting this package as outdated make sure there is indeed a new version for Linux Desktop. You can have a look at the "Stable updates" tag in Release blog for this.
You can also run this command to obtain the version string for the latest chrome version:
Do not report updates for ChromeOS, Android or other platforms stable versions as updates here.