Repeated input in PTY with SSH when a native process runs a Cygwin process that consumes input

Repeated input in PTY with SSH when a native process runs a Cygwin process that consumes input

[Andrew Ng]

Hi,

There appears to be an issue with repeated input in PTY (with pcon enabled)
when using SSH. If I run MinGW git from the bash shell via ssh, and git
then executes Cygwin less, e.g. for git log, then any input to less is
repeated as input after git exits. So typically, you will end up with some
number of space characters and 'q' in your terminal.

I have also seen this same behaviour in MSYS2 with pcon enabled.

Both Cygwin and MSYS2 are running 3.1.4.

Thank you,
Andrew Ng

Re: Repeated input in PTY with SSH when a native process runs a Cygwin process that consumes input

[Takashi Yano]

On Sat, 30 May 2020 15:46:39 +0100
Andrew Ng via Cygwin <cygwin@cygwin.com> wrote:
> Hi,
> 
> There appears to be an issue with repeated input in PTY (with pcon enabled)
> when using SSH. If I run MinGW git from the bash shell via ssh, and git
> then executes Cygwin less, e.g. for git log, then any input to less is
> repeated as input after git exits. So typically, you will end up with some
> number of space characters and 'q' in your terminal.
> 
> I have also seen this same behaviour in MSYS2 with pcon enabled.
> 
> Both Cygwin and MSYS2 are running 3.1.4.

Thanks for the report. I looked into this problem and found
the cause. Let me consider how to fix.

-- 
Takashi Yano <takashi.yano@nifty.ne.jp>

Re: Repeated input in PTY with SSH when a native process runs a Cygwin process that consumes input

[Takashi Yano]

Hi Andrew,

On Sun, 31 May 2020 01:06:28 +0900
Takashi Yano via Cygwin <cygwin@cygwin.com> wrote:
> On Sat, 30 May 2020 15:46:39 +0100
> Andrew Ng via Cygwin <cygwin@cygwin.com> wrote:
> > Hi,
> > 
> > There appears to be an issue with repeated input in PTY (with pcon enabled)
> > when using SSH. If I run MinGW git from the bash shell via ssh, and git
> > then executes Cygwin less, e.g. for git log, then any input to less is
> > repeated as input after git exits. So typically, you will end up with some
> > number of space characters and 'q' in your terminal.
> > 
> > I have also seen this same behaviour in MSYS2 with pcon enabled.
> > 
> > Both Cygwin and MSYS2 are running 3.1.4.
> 
> Thanks for the report. I looked into this problem and found
> the cause. Let me consider how to fix.

The developers snapshot, in which the issue has been fixed,
is ready. Could you please test?

-- 
Takashi Yano <takashi.yano@nifty.ne.jp>

Re: Repeated input in PTY with SSH when a native process runs a Cygwin process that consumes input

[Andrew Ng]

Hi Takashi,

> The developers snapshot, in which the issue has been fixed,
> is ready. Could you please test?

Thanks for the fast response.

I have tested the snapshot and can confirm that it fixes the issue. I
have also applied the same fix to MSYS2 and confirmed the issue is
fixed there too.

I did take a look at the code when I came across this issue and I
noticed a few things that looked a bit odd. I noticed that the
following functions appear to make use of rabuf () without
consideration for raixget ():

  fhandler_termios::eat_readahead
  fhandler_pty_master::accept_input
  fhandler_pty_master::transfer_input_to_pcon

Also fhandler_pty_slave::eat_readahead does 'eols' processing without
any guard on is_line_input (), which seems to differ from the code
pattern seen elsewhere.

I don't know if these are important, but thought I should mention
them, just in case.

Thanks,
Andrew

Re: Repeated input in PTY with SSH when a native process runs a Cygwin process that consumes input

[Takashi Yano]

On Sun, 31 May 2020 13:00:51 +0100
Andrew Ng <anng.sw@gmail.com> wrote:
> > The developers snapshot, in which the issue has been fixed,
> > is ready. Could you please test?
> 
> Thanks for the fast response.
> 
> I have tested the snapshot and can confirm that it fixes the issue. I
> have also applied the same fix to MSYS2 and confirmed the issue is
> fixed there too.

Thanks for testing.

> I did take a look at the code when I came across this issue and I
> noticed a few things that looked a bit odd. I noticed that the
> following functions appear to make use of rabuf () without
> consideration for raixget ():
> 
>   fhandler_termios::eat_readahead
>   fhandler_pty_master::accept_input
>   fhandler_pty_master::transfer_input_to_pcon
> 
> Also fhandler_pty_slave::eat_readahead does 'eols' processing without
> any guard on is_line_input (), which seems to differ from the code
> pattern seen elsewhere.
> 
> I don't know if these are important, but thought I should mention
> them, just in case.

Thank you for your advice. I will check if these can cause any
problem.

As for pty master, get_readahead() is not used at all, so raixget()
is always 0. So pty master code does not cause problem.

However, it is certainly dangerous. I will consider how these
should be.

-- 
Takashi Yano <takashi.yano@nifty.ne.jp>