A more-conservative workaround for Cygwin SIGCHLD issues.

* callproc.c (Fcall_process):
* process.c (create_process):
Make sure SIGCHLD is caught before we fork,
since Emacs startup no arranges to catch SIGCHLD.
* process.c (lib_child_handler): Initialize to null, not to
dummy_handler.
(catch_child_signal): Allow self to be called lazily.
Do nothing if it's already been called.
Assume caller has blocked SIGCHLD (all callers do now).
* emacs.c (main): Do not catch SIGCHLD here; defer it until
just before it's really needed.
* nsterm.m (ns_term_init): No need to re-catch SIGCHLD here,
since it hasn't been caught yet.

Fixes: debbugs:14569

1 files changed, 1 insertions, 0 deletions

diff --git a/src/callproc.c b/src/callproc.c
index f0aa8222342..7db984fa71c 100644
--- a/src/callproc.c
+++ b/src/callproc.c
@@ -613,6 +613,7 @@ usage: (call-process PROGRAM &optional INFILE DESTINATION DISPLAY &rest ARGS)  *
 
     block_input ();
     block_child_signal ();
+    catch_child_signal ();
 
 #ifdef WINDOWSNT
     pid = child_setup (filefd, fd1, fd_error, new_argv, 0, current_dir);