Re: fixing memory leaks before the pretest (was: Update on the Emacs release schedule?)

[YAMAMOTO Mitsuharu]

>>>>> On Sat, 7 Jan 2012 17:57:38 +0100, Carsten Mattner <address@hidden> said:

> 2012/1/7 Ted Zlatanov <address@hidden>:
>> On Sat, 07 Jan 2012 15:10:38 +0800 Chong Yidong <address@hidden>
>> wrote:
>> 
CY> As for the code, there are still a number of issues that need more
CY> attention, most prominently the mysterious memory leak(s) that may
CY> or may not involve Gnus and/or GnuTLS and/or Mac OS X.
>> 
>> I am seeing memory growth on GNU/Linux with Gnus and GnuTLS that I
>> don't see otherwise without Gnus, so it's faintly possible GnuTLS
>> is not the determining factor.  I have gone over the gnutls.c code
>> and don't see where the GnuTLS glue could be leaking.  If it is,
>> I'll need a tool like Valgrind to help me, and last time I tried
>> that, the reports were not helpful to me (too much data, not enough
>> leading back to GnuTLS).  I spent 2 days on this last week and
>> meant to bring it up this week, actually (the discussion about
>> GnuTLS on W32 sort of distracted me :)
>> 
>> Maybe someone who actually knows how to use Valgrind could help me
>> or try to find the leaks themselves?

> Tried LLVM AddressSanitizer? It's supposed to be a low-impact
> compile flag.  http://clang.llvm.org/docs/AddressSanitizer.html
> From the documentation I'm not sure it's useful for finding leaks.

I suggested a possible way to analyze heap usage on Mac OS X only
using some commands that bundled with the standard developer tools:

  http://lists.gnu.org/archive/html/emacs-devel/2011-12/msg00552.html .

Has anybody who are seeing memory issues on Mac OS X tried this?  At
least we can make sure whether there are some bogus root references or
freed memory is not returned to the system.

                                     YAMAMOTO Mitsuharu
                                address@hidden