If process() failed and the signal remains in the queue, the most possible reason is that the target thread already armed by another signal and do not handle it yet. With this patch, to increase the chance of handling it in the other threads, call yield() before retrying process().
Addresses: https://cygwin.com/pipermail/cygwin/2024-November/256744.html Fixes: e10f822a2b39 ("Cygwin: signal: Handle queued signal without explicit __SIGFLUSH") Reported-by: Christian Franke <christian.fra...@t-online.de> Reviewed-by: Signed-off-by: Takashi Yano <takashi.y...@nifty.ne.jp> --- winsup/cygwin/sigproc.cc | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/winsup/cygwin/sigproc.cc b/winsup/cygwin/sigproc.cc index 4c557f048..7e02e61f7 100644 --- a/winsup/cygwin/sigproc.cc +++ b/winsup/cygwin/sigproc.cc @@ -1342,7 +1342,10 @@ wait_sig (VOID *) pack.si.si_signo = __SIGFLUSH; else if (sigq.start.next && PeekNamedPipe (my_readsig, NULL, 0, NULL, &nb, NULL) && !nb) - pack.si.si_signo = __SIGFLUSH; + { + yield (); + pack.si.si_signo = __SIGFLUSH; + } else if (!ReadFile (my_readsig, &pack, sizeof (pack), &nb, NULL)) Sleep (INFINITE); /* Assume were exiting. Never exit this thread */ else if (nb != sizeof (pack) || !pack.si.si_signo) -- 2.45.1
