M-x shell breakage with tcsh

[Jerry James]

I've received a report that M-x shell is broken for a tcsh user on Fedora:

https://bugzilla.redhat.com/show_bug.cgi?id=1222897

Playing with it a bit, I discovered that M-x gdb is also somewhat
broken, in the same way. The subprocesses are calling tcgetsid() on
the pipes that connect them to the XEmacs process, and are getting
ENOTTY responses. If anybody understands how that code works and
could give some guidance on how to resolve the issue, your help would
be much appreciated.

The user reports that this problem does NOT happen with Fedora 21
XEmacs, which is a build of mercurial commit acf1c26e3019, but it DOES
happen with Fedora 22 XEmacs, which is a build of mercurial commit
23178aa71f8b.

Thanks,

[Henry S. Thompson]

Could be related to these threads:

http://list-archive.xemacs.org/pipermail/xemacs-beta/2014-January/024736.html
http://list-archive.xemacs.org/pipermail/xemacs-beta/2014-January/024741.html
http://list-archive.xemacs.org/pipermail/xemacs-beta/2014-May/024848.html
http://list-archive.xemacs.org/pipermail/xemacs-beta/2015-January/025451.html

Try this workaround:

(setq process-connection-type nil)

Just checking commit ids wrt the patch that fixed this, perhaps not
for all platforms -- it was

changeset: 5882:4949ccab25f1
user: Vin Shelton <***@xemacs.org>
date: Tue Mar 24 15:50:43 2015 -0400
files: src/ChangeLog src/process-unix.c
description:
Get M-x shell working on cygwin

Which does indeed fall _between_ the two points you identify as
working vs. not working. . .

ht