bug#1504: 23.0.60; emacs daemon leaks memory

bug#1504: 23.0.60; emacs daemon leaks memory

[Markus Triska]

When I do:

   $ emacs -Q --daemon

followed by:

   $ while true; do emacsclient -t -e "(save-buffers-kill-terminal)"; done

then the daemon apparently consumes increasingly more memory.

In GNU Emacs 23.0.60.9 (i386-apple-darwin8.11.1, GTK+ Version 2.12.9)
 of 2008-11-28 on mt-computer.local
Windowing system distributor `The XFree86 Project, Inc', version 11.0.40400000
Important settings:
  value of $LC_ALL: nil
  value of $LC_COLLATE: nil
  value of $LC_CTYPE: nil
  value of $LC_MESSAGES: nil
  value of $LC_MONETARY: nil
  value of $LC_NUMERIC: nil
  value of $LC_TIME: nil
  value of $LANG: en_GB.UTF-8
  value of $XMODIFIERS: nil
  locale-coding-system: nil
  default-enable-multibyte-characters: t

bug#1504: marked as done (23.0.60; emacs daemon leaks memory)

[Emacs bug Tracking System]

[-- Attachment #1: Type: text/plain, Size: 836 bytes --]


Your message dated Thu, 11 Dec 2008 15:50:13 -0500
with message-id <871vwecu7u.fsf@stupidchicken.com>
and subject line Re: 23.0.60; emacs daemon leaks memory
has caused the Emacs bug report #1504,
regarding 23.0.60; emacs daemon leaks memory
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact don@donarmstrong.com
immediately.)


-- 
1504: http://emacsbugs.donarmstrong.com/cgi-bin/bugreport.cgi?bug=1504
Emacs Bug Tracking System
Contact don@donarmstrong.com with problems

[-- Attachment #2: Type: message/rfc822, Size: 2712 bytes --]

From: Markus Triska <markus.triska@gmx.at>
To: emacs-pretest-bug@gnu.org
Subject: 23.0.60; emacs daemon leaks memory
Date: Sat,  6 Dec 2008 12:20:09 +0100 (CET)
Message-ID: <20081206112009.E93A0BD9391@mt-computer.local>


When I do:

   $ emacs -Q --daemon

followed by:

   $ while true; do emacsclient -t -e "(save-buffers-kill-terminal)"; done

then the daemon apparently consumes increasingly more memory.

In GNU Emacs 23.0.60.9 (i386-apple-darwin8.11.1, GTK+ Version 2.12.9)
 of 2008-11-28 on mt-computer.local
Windowing system distributor `The XFree86 Project, Inc', version 11.0.40400000
Important settings:
  value of $LC_ALL: nil
  value of $LC_COLLATE: nil
  value of $LC_CTYPE: nil
  value of $LC_MESSAGES: nil
  value of $LC_MONETARY: nil
  value of $LC_NUMERIC: nil
  value of $LC_TIME: nil
  value of $LANG: en_GB.UTF-8
  value of $XMODIFIERS: nil
  locale-coding-system: nil
  default-enable-multibyte-characters: t




[-- Attachment #3: Type: message/rfc822, Size: 1724 bytes --]

From: Chong Yidong <cyd@stupidchicken.com>
To: Markus Triska <markus.triska@gmx.at>
Cc: 1504-done@emacsbugs.donarmstrong.com
Subject: Re: 23.0.60; emacs daemon leaks memory
Date: Thu, 11 Dec 2008 15:50:13 -0500
Message-ID: <871vwecu7u.fsf@stupidchicken.com>

Aha: Emacs wasn't freeing the frame face cache on tty frames.  Fixed.

This, plus a couple of other leaks in term.c that I fixed earlier this
week, accounts for about 95% of the leakage you reported.

There remains a smaller leak occurring in Lisp: see

http://lists.gnu.org/archive/html/emacs-devel/2008-12/msg00443.html

I'll continue looking into that, but it's less urgent because that leak
is much, much slower.  In the meantime, I'm closing the bug.  Thanks for
reporting.

bug#1504: closed by Chong Yidong <cyd@stupidchicken.com> (Re: 23.0.60; emacs daemon leaks memory)

[Markus Triska]

> it's less urgent because that leak is much, much slower.

Serious emacsclient testing can only start when the leak is completely
fixed: I'm trying to trace down a different problem (#1125) which only
seems to happen after entering and exiting emacsclient occasionally.
Still, thank you for the current fixes - I can now run the test cases
for a few minutes instead of only seconds, which is already better.

Also, when I use emacsclient with "-c" instead of "-t" in the test case,
there's a leak that seems to be as severe as before your patches.