[Bug 57097] New: Keyboard inputs for special characters on non-english keyboards are switched in winewayland
https://bugs.winehq.org/show_bug.cgi?id=57097 Bug ID: 57097 Summary: Keyboard inputs for special characters on non-english keyboards are switched in winewayland Product: Wine Version: 9.16 Hardware: x86-64 OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: winewayland Assignee: wine-bugs(a)winehq.org Reporter: sutupud(a)yahoo.com Distribution: --- I noticed that if I use a german keyboard layout, the key mapping seems to be off. I'm using kwin_wayland (but also tested on sway, same result). The regular lower keys are ok, but example, ö produces ü, # produces ä, etc. Looking at dlls/winewayland.drv/wayland_keyboard.c, I saw that there are different lookup tables for different languages (scan2vk_qwerty, scan2vk_azerty. scan2vk_qwertz, ...), so my first quess was that something might be switched there. But that doesn't really seem the case, changing it to always use `scan2vk_qwertz ` regardless of `lang` makes it work. I don't really have a very good understanding of how this is supposed to work, but does the appliaction really need to do such a translation? Isn't that already handled by the layout being passed in? At least that would explain why it works when doing no such lookup at all. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=57097 Ken Sharp <imwellcushtymelike(a)gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Keywords| |download, localization, | |source -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=57097 --- Comment #1 from mata <sutupud(a)yahoo.com> --- Created attachment 77568 --> https://bugs.winehq.org/attachment.cgi?id=77568 ignore switch that breaks keysyms -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=57097 --- Comment #2 from mata <sutupud(a)yahoo.com> --- Tested again in wine-10.0rc1, issue is still present. Skipping the switch that selects the lookup table still works around it. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=57097 Fell <fell(a)fellr.net> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |fell(a)fellr.net --- Comment #3 from Fell <fell(a)fellr.net> --- This is still present in Wine 10.1 Steps to reproduce: 1. Start a Wayland session in KDE Plasma. 2. Set the keyboard layout to something nonstandard, like German QWERTZ. 3. Run `DISPLAY= wine notepad` to run Notepad using the native Wayland driver. 4. Begin typing some of the characters mentioned below. 5. Observe, that some of the produced characters to not correspond to what is typed. Some observations: all of the following are mapped correctly: !"§$%&/()=¹²³{[]}|<>@:. +-*/ on the numpad are mapped correctly A-X are all mapped correctly, Y and Z are swapped All digits (numpad and not numpad) are mapped correctly typing Ä produces Ö typing Ö produces Ü typing Ü produces ? (actual question mark) typing ° produces * typing ^ produces + typing + (not numpad) produces nothing at all (the cursor doesn't advance) typing * (not numpad) produces ` typing ` or ´ produces nothing at all typing ? produces _ typing ß procuces - typing € produces 0x80, which is correct in Windows-1252 encoding, but is rendered as a vertical bar, slightly shorter than | (this might be a whole different bug) System Information: Wine Version: 10.1 (from official Arch Linux repository) Operating System: Arch Linux KDE Plasma Version: 6.3.0 KDE Frameworks Version: 6.10.0 Qt Version: 6.8.2 Kernel Version: 6.13.2-arch1-1 (64-bit) Graphics Platform: Wayland Processors: 24 × AMD Ryzen 9 3900X 12-Core Processor Memory: 62.7 GiB of RAM Graphics Processor: NVIDIA GeForce RTX 2080 Ti -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=57097 --- Comment #4 from Fell <fell(a)fellr.net> --- This is still present in Wine 10.1 Steps to reproduce: 1. Start a Wayland session in KDE Plasma. 2. Set the keyboard layout to something nonstandard, like German QWERTZ. 3. Run `DISPLAY= wine notepad` to run Notepad using the native Wayland driver. 4. Begin typing some of the characters mentioned below. 5. Observe, that some of the produced characters to not correspond to what is typed. Some observations: all of the following are mapped correctly: !"§$%&/()=¹²³{[]}|<>@:. +-*/ on the numpad are mapped correctly A-X are all mapped correctly, Y and Z are swapped All digits (numpad and not numpad) are mapped correctly typing Ä produces Ö typing Ö produces Ü typing Ü produces ? (actual question mark) typing ° produces * typing ^ produces + typing + (not numpad) produces nothing at all (the cursor doesn't advance) typing * (not numpad) produces ` typing ` or ´ produces nothing at all typing ? produces _ typing ß procuces - typing € produces 0x80, which is correct in Windows-1252 encoding, but is rendered as a vertical bar, slightly shorter than | (this might be a whole different bug) System Information: Wine Version: 10.1 (from official Arch Linux repository) Operating System: Arch Linux KDE Plasma Version: 6.3.0 KDE Frameworks Version: 6.10.0 Qt Version: 6.8.2 Kernel Version: 6.13.2-arch1-1 (64-bit) Graphics Platform: Wayland Processors: 24 × AMD Ryzen 9 3900X 12-Core Processor Memory: 62.7 GiB of RAM Graphics Processor: NVIDIA GeForce RTX 2080 Ti -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=57097 --- Comment #5 from Fell <fell(a)fellr.net> --- Sorry for the double posting earlier. That was a mistake. I can now also confirm that the workaround by @mata does indeed still fix the issue. The code in question seems to handle specifically French, German, Swiss and various "dvorak" layouts, which seems odd. ``` switch (lang) { case MAKELANGID(LANG_FRENCH, SUBLANG_DEFAULT): scan2vk = scan2vk_azerty; break; case MAKELANGID(LANG_GERMAN, SUBLANG_DEFAULT): scan2vk = scan2vk_qwertz; break; case MAKELANGID(LANG_GERMAN, SUBLANG_GERMAN_SWISS): scan2vk = scan2vk_qwertz; break; default: scan2vk = scan2vk_qwerty; break; } if (strstr(xkb_layout, "dvorak")) scan2vk = scan2vk_dvorak; ``` It was added in this commit: https://gitlab.winehq.org/wine/wine/-/commit/b2d16b920431820121c849e8bbc462c... Could this have been a workaround for a potentially wrong keyboard handling in some other Wayland compositor than KWin? Could someone try this in other desktop environments? -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=57097 Mershl <mweires(a)googlemail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |mweires(a)googlemail.com -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
https://bugs.winehq.org/show_bug.cgi?id=57097 --- Comment #6 from mata <sutupud(a)yahoo.com> --- Still no change in wine 10.4 I've tried to make some sense of those remappings, it seems to try to fix the layout internally, but the xcb layout should already do that. But even with that assumption it doesn't seem to work, e.g. it doesn't really translate a qwerty layout to qwertz if the lookup table is selected accordingly. And if that was required, it would need to have similar tables for any other potential layout. The compositors with which I tried it are kwin_wayland, weston and sway - all with the same result. Another minor problem I saw is that the numlock state can't be toggled. A while ago I remember it being stuck in off state, now it's always on. Not sure if that's somewhow related. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
http://bugs.winehq.org/show_bug.cgi?id=57097 Rémi Bernon <rbernon(a)codeweavers.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |rbernon(a)codeweavers.com --- Comment #7 from Rémi Bernon <rbernon(a)codeweavers.com> --- The switch is correct and it is there to set the proper Win32 scan code to vkey tables depending on the keyboard layout language. The issue with winewayland layouts seems to be coming from a mismatch in keyboard layout selection: Whenever winewayland receives keymap change notification, we change the focused window active layout by sending it an WM_INPUTLANGCHANGEREQUEST window message. This message, then calls NtUserActivateKeyboardLayout with the selected layout. This sends a WM_INPUTLANGCHANGE message to the focused window, changing its thread info kbd_layout to the active layout. When winewayland receives a keyboard input, we send the keyboard input as a raw scancode, which is supposed to later be translated to the proper vkey according to the active keyboard layout. However, that conversion happens in send_hardware_message which is called directly from NtUserSendHardwareMessage, which is itself called directly from the winewayland keyboard input handler, in the context of the winewayland event thread, which is an internal thread and different from any window thread. That thread never changed its active layout and still uses the default keyboard layout, which is always matching the user locale (may be changed with LC_ALL env var). The keyboard layout problem can be worked around by setting LC_ALL= to the language of the desired keyboard layout. It then could probably be fixed by calling NtUserActivateKeyboardLayout in the context of the winewayland thread, in addition to sending the WM_INPUTLANGCHANGEREQUEST message, although that message could also be hooked by application to refuse such keyboard layout activation and that should be handled. Alternatively, the hardware message scan code -> vkey conversion could be moved to the receiving window thread, which has the right keyboard layout activated, but that would require some changes in wineserver, as it currently relies on keyboard hardware message vkey being set. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
http://bugs.winehq.org/show_bug.cgi?id=57097 --- Comment #8 from Mershl <mweires(a)googlemail.com> --- can confirm the proposed LC_ALL=de_DE of Rémi works around the issue. tested on german keyboardlayout and winewayland 10.5 -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
http://bugs.winehq.org/show_bug.cgi?id=57097 Bungeefan <severin.hamader(a)yahoo.de> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |severin.hamader(a)yahoo.de -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
http://bugs.winehq.org/show_bug.cgi?id=57097 --- Comment #9 from mata <sutupud(a)yahoo.com> --- Setting LC_ALL can barely be considered a workaround. The LC_* sections should not really have anything to do with the keyboard layout. Or which LC_* subsection concretely needs to be set as to not override just this, without affecting anything else. I can only say that when I disable that switch, it works for me in any locale and keyboard layout combination. With that switch in place it doesn't. So something clearly seems to be wrong with that logic. https://bugs.winehq.org/show_bug.cgi?id=58247 also might be related. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
http://bugs.winehq.org/show_bug.cgi?id=57097 Totto <tobiausgais(a)gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |tobiausgais(a)gmail.com -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
http://bugs.winehq.org/show_bug.cgi?id=57097 GoodAlchemist <severin.hamader@yahoo.de> changed: What |Removed |Added ---------------------------------------------------------------------------- CC|severin.hamader@yahoo.de | -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.
participants (2)
-
WineHQ Bugzilla -
WineHQ Bugzilla