Hi Thomas, On Jan 30 21:59, Thomas Wolff via Cygwin wrote: > > Am 28.01.2025 um 09:56 schrieb Corinna Vinschen via Cygwin: > > On Jan 28 00:50, Thomas Wolff via Cygwin wrote: > > > Am 27.01.2025 um 12:17 schrieb Takashi Yano via Cygwin: > > > > Hi Thomas, > > > > > > > > A few days ago, Corinna asked me to check a problem of TTY. > > > > The problem is as follows. > > > > > > > > Reproduce steps: > > > > (1) Open mintty. > > > > (2) Open another mintty. > > > > (3) Place the second mintty window over the first one. > > > > (4) Hold ctrl key down. > > > > (5) Press 'd' key while holding ctrl key. The second mintty > > > > window will be closed. Keep ctrl key still hold down. > > > > (6) Now the first mintty window is focused. Then press 'd' key > > > > while holding ctrl key down. > > > > (7) The first mintty window will not logout but displays 'd'. > > > > > > > > I checked the pty and found the pty in the first mintty > > > > receives just 'd' but not Ctrl-D from mintty. In other words, > > > > fhandler_pty_master::write() is called with 'd'. So I suspect it > > > > might be a problem of mintty. Therefore, I tried old version > > > > of mintty. The results are: > > > > > > > > ... > > > > > > > Hi Takashi, > > > sorry this was attributed to pty, it is an issue of mintty I was aware > > > of which appeared in 3.7.2. > > > More precisely, it slipped in with > > > be73970877a99548aeeab60a2572ffb04b695066 "revise AltGr handling to > > > support flexible right-Alt+left-Ctrl combinations (#1266)", as a > > > trade-off for what I meant to be a final fix for control modifier > > > handling. I guess I hadn't considered it serious enough to reopen the > > > issue for another workaround. To be reconsidered... > > I just had a look into that commit and saw the description of a hidden > > option OldAltGrDetection right at the start. Looks like > > OldAltGrDetection=2 is a temporary workaround for the problem. What's > > the drawback of using it? > I've uploaded a fix to the github repository and tried to find an answer > to your question but couldn't yet... > Some background information: root cause of the trouble is Windows' > insane handling of the AltGr modifier and its failure to assign a > distinct virtual keycode to it. Instead it provides left-Control and > right-Alt key events, both with the same timestamp. That alone is a bit > tricky to detect as both Control and Alt are modifiers on their own. > Worse, software handling of AltGr detection or (in the case of remote > desktop software) generation often deviates from this scheme and even > Windows' own onscreen keyboard is buggy about this. Mintty has a number > of workarounds to handle that. Also, basically in a terminal emulator > the expectation is to handle all of Control, Alt etc and also their > combinations properly. For a example, on a keyboard where AltGr+q is > '@', Control+AltGr+q should be ^@, Alt+AltGr+q should be ESC @, and > Control+Alt+AltGr+q should be ESC ^@. All this is accomplished by > mintty. Together with the workarounds mentioned above, this is a > delicate area and I hope the fix found now interworks well with all such > software. > Thomas
Thanks for your description. I'm running mintty with OldAltGrDetection=2 for a couple of days now and didn't have any problems. I assume the complicated detection cases you're working on don't show up in my workflow. Corinna -- Problem reports: https://cygwin.com/problems.html FAQ: https://cygwin.com/faq/ Documentation: https://cygwin.com/docs.html Unsubscribe info: https://cygwin.com/ml/#unsubscribe-simple
