Re: Debugging memory leaks/stale references

[Simon Josefsson <jas@extundo.com>]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> while they are traversed.  Unfortunately, no major leak turned up,
>> just a few thousand cons cells which can't cause the massive leak I
>> see (in the order of dozens of megabytes after entering/exiting a
>> large NNTP group with Gnus).
>
> I think it would be worthwhile to start with the basic info returned by
> `garbage-collect' and such stuff.  See `memory-usage.el' below.
> [ I thought you already tried that, but the few lines above make me think
> that you don't even know whether the extra MBs are made up of cons cells, or
> strings, or non-elisp-data, or ...]

Thanks, memory-usage seem useful.  The memory growth caused by Gnus
seem to escape it though.  This is a fairly fresh emacs (I'm using
Gnus, programming C and reading documentation, but nothing fancy):

20807 jas       15   0  365m 185m  17m S  2.7 18.3  89:12.64 emacs

But there is no indication of where the memory is used (see below).

Perhaps the "top" statistics aren't reliable, but the fact remains
that the sluggishness I feel appear to be correlated to high memory
usage according to "top".

Hm.  Is it possible to use 'gdb' to find out where all memory goes?
Seems like it should have the capability to do it, at least.

(FWIW, the above emacs is not sluggish.)

Garbage collection stats:
((1130529 . 100249) (60991 . 170) (8108 . 5450) 3508516 1049895 (322 . 551) (34955 . 28764) (143989 . 25103))

 =>	9846224 bytes in cons cells
	1467864 bytes in symbols
	271160 bytes in markers
	3508516 bytes of string chars
	4199580 bytes of vector slots
	10476 bytes in floats
	1784132 bytes in intervals

Total bytes in lisp objects (not counting string and vector headers): 17938267

Buffer ralloc memory usage:
147 buffers
5564167 bytes total (66581 in gaps)
      Size	Gap	Name

   1843610	2000	 *DOC*
    524443	2000	snprintf_2.2.tar.gz
    467401	2000	*Messages*
    460876	2000	 *Gnus Backlog*
...