Re: Emacs daemon dies at Xorg crash

[Dan Nicolaescu <dann@ics.uci.edu>]

Ulrich Mueller <ulm@gentoo.org> writes:

  > >>>>> On Thu, 18 Dec 2008, Dan Nicolaescu wrote:
  > 
  > >> > Can you try the patch at:
  > >> > http://permalink.gmane.org/gmane.emacs.bugs/22201
  > >> 
  > >> Tested and Emacs doesn't die anymore. I find the "Connection lost ..."
  > >> message in the *Messages* buffer after reconnecting.
  > 
  > > IMO that patch is good, but it fixes the effect in this case, not the
  > > cause.
  > 
  > Hm, could you please elaborate what you mean by "effect" and "cause"
  > here?

Emacs dies because something tries to call message while no frame is
open.  This is the "effect".  The "cause" is whatever causes that
message to be written, it probably should not happen.  And indeed, even
for you, it happens only for some specific window managers.


  > 
  > >> #0  Fsignal (error_symbol=0x84863b9, data=0x874d62d) at eval.c:1640
  > >> #1  0x0824c780 in xsignal (error_symbol=0x84863b9, data=0x874d62d) at eval.c:1772
  > >> #2  0x081f8844 in report_file_error (string=0x82e8afa "make client process failed", data=0x874d675) at fileio.c:277
  > >> #3  0x0829a4bd in Fmake_network_process (nargs=0xa, args=0x7f877cd0) at process.c:3632
  > 
  > This error has nothing to do with the X connection problem, right?

I don't know, you need to find out why we get here sometimes, but not
when using other WMs.

  > > Can you please check why the above function fails?
  > 
  > server-start gets called from kill-emacs-hook. It first deletes the
  > server-process, and then calls server-running-p, which unsuccessfully

This sounds strange... 

  > tries to connect to the server's socket (server.el, revision 1.177):
  > 
  >       (when server-process
  >         ;; kill it dead!
  >         (ignore-errors (delete-process server-process)))
  >       ;; Delete the socket files made by previous server invocations.
  >       (if (not (eq t (server-running-p server-name)))
  > 
  > However, please note that make-network-process in server-running-p is
  > inside a condition-case. (Which doesn't prevent it from hitting my
  > gdb breakpoint in Fsignal.)
  > 
  > > Can you also revert server.el to version 1.175 and see if that still
  > > causes problems?
  > 
  > With 1.175 it doesn't hit Fsignal. But I'm not sure if there really is
  > a problem in 1.177.