Race condition hangs on multiple mintty/tcsh? Brad Wetmore

Thomas Wolff towo@towo.net
Thu Aug 6 11:46:08 GMT 2020


Am 06.08.2020 um 01:23 schrieb Kevin Schnitzius via Cygwin:
> On Wednesday, August 5, 2020, 06:56:48 PM EDT, Thomas Wolff <towo@towo.net> 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.
Does anybody familiar with pty/select or recent changes have any idea?
Thomas


More information about the Cygwin mailing list