[PATCH 9/9] Cygwin: signal: Fix a short period of deadlock

[Corinna Vinschen]

On Nov 29 20:59, Takashi Yano wrote:
> The main thread waits for the sig thread to read the signal pipe by
> calling Sleep(10) if writing to the signal pipe has failed. However,
> if the signal thread waiting for another signal being handled in the
> main thread, the sig thread does not read the signal pipe. To avoid
> such a situation, this patch replaces Sleep(10) to cygwait().
> 
> Addresses: https://cygwin.com/pipermail/cygwin/2024-November/256744.html
> Fixes: 6f05b327678f ("(sig_send): Retry WriteFiles which fail when there is no error but packbytes have not been sent.")
> Reported-by: Christian Franke <Christian.Franke@t-online.de>
> Reviewed-by:
> Signed-off-by: Takashi Yano <takashi.yano@nifty.ne.jp>
> ---
>  winsup/cygwin/sigproc.cc | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/winsup/cygwin/sigproc.cc b/winsup/cygwin/sigproc.cc
> index 8c788bd20..4c557f048 100644
> --- a/winsup/cygwin/sigproc.cc
> +++ b/winsup/cygwin/sigproc.cc
> @@ -741,7 +741,8 @@ sig_send (_pinfo *p, siginfo_t& si, _cygtls *tls)
>        res = WriteFile (sendsig, leader, packsize, &nb, NULL);
>        if (!res || packsize == nb)
>  	break;
> -      Sleep (10);
> +      if (cygwait (NULL, 10, cw_sig_eintr) == WAIT_SIGNALED)
> +	_my_tls.call_signal_handler ();
>        res = 0;
>      }
>  
> -- 
> 2.45.1

LGTM.


Thanks,
Corinna