Hi Thomas, On Thu, 6 Aug 2020 15:31:24 +0200 Thomas Wolff wrote: > Am 06.08.2020 um 13:46 schrieb Thomas Wolff: > > Am 06.08.2020 um 01:23 schrieb Kevin Schnitzius via Cygwin: > >> On Wednesday, August 5, 2020, 06:56:48 PM EDT, Thomas Wolff > >> wrote: > >>> Am 04.08.2020 um 12:02 schrieb Thomas Wolff: > >>>> Am 04.08.2020 um 00:13 schrieb Brad Wetmore via Cygwin: > >>>>> Hi, > >>>>> > >>>>> I generally kick off multiple (10) mintty sessions, and place them > >>>>> around the screen. > >>>>> > >>>>> C:/cygwin64/bin/mintty.exe -i /Cygwin-Terminal.ico \ > >>>> --position @3 --position 120,0 --size 80x71 /bin/tcsh & > >>>>> C:/cygwin64/bin/mintty.exe -i /Cygwin-Terminal.ico \ > >>>>> --position @3 --position 715,0 --size 80x45 /bin/tcsh & > >>>>> C:/cygwin64/bin/mintty.exe -i /Cygwin-Terminal.ico \ > >>>>> --position @3 --position 715,660 --size 80x24 /bin/tcsh & > >>>>> > >>>>> Within the last 6 months or so, about 2-3 of them would hang and > >>>>> either mintty/tcsh would not start. I put a "sleep 1" in between each > >>>>> invocation and that seemed to take care of it. > >>>>> > >>>>> With the latest cygwin update, about 8 of them just hang even with > >>>>> the sleep 1. I put in a "sleep 2", and now everything is coming up > >>>>> again. > >>>>> > >>>>> Not sure if this is a mintty or tcsh issue, but just wondering if > >>>>> others are seeing this before I start trying to debug this. > >>>> I can reproduce such behaviour with /bin/bash (easy cross-check), and > >>>> in fact the shell is running in that case (easy test via `echo > > >>>> .log`), so I have a vague and unpleasant suspicion it might in fact be > >>>> related to mintty although I have no idea how that would happen. To be > >>>> analysed. > >>> When in this state, on the pty from which mintty receives child process > >>> output, select() does not report a ready for reading condition; > >>> could it > >>> be related to the recent poll/select patch? I could not reproduce it in > >>> cygwin 3.0.7. > >> for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ; do > >> mintty  -i /Cygwin-Terminal.ico  --position @3 --position 120,$i > >> --size 80x24 /bin/bash & done > >> > >> This does not fail for me.  Nor does it fail from cmd or powershell > >> using a script. > >> > >> I tried 3.1.4 and 3.1.6 on Windows 10.  I tried up to 100 instances > >> of of mintty... > > Thanks for testing. It is in fact hard to reproduce, maybe also > > depending on system load (speculating). > > I could reproduce one case of one of three terminals being > > unresponsive also with xterm. > And it also happens if I drop select() from mintty (and use just > non-blocking read()). > > Does anybody familiar with pty/select or recent changes have any idea? I looked into this problem. After much struggle, I think I have found a workaround for this issue. I am not sure why this solves the issue at all, however, this works for me. Could you please test a patch attached? -- Takashi Yano