Re: Malloc Problem in glibc 2.2

[Andreas Jaeger <aj at suse dot de>]

>>>>> Wolfram Gloger writes:

Wolfram> Hello,
>> Forget the rlimit issues, I've just used them to illustrate a problem
>> I had with malloc.

Wolfram> OK.

>> Do I understand you correctly that you changed the meanings of the
>> environment variables MALLOC_MMAP_THRESHOLD_ and MALLOC_MMAP_MAX_ (and
>> the corresponding mallopt options) in a non compatible way - which
>> violates the documentation?

Wolfram> The effect of MALLOC_MMAP_THRESHOLD_ isn't changed.

It has been changed in the way that mmap is now also used for smaller
regions if sbrk fails.

Wolfram> I just want the default behaviour to be `address space completely
Wolfram> allocatable', that's why I changed the meaning of MALLOC_MMAP_MAX_.

>> Playing around with MALLOC_MMAP_MAX I noticed that it's impossible to
>> switch off mmap completly.  Do you really define this as a "feature"
>> of glibc 2.2?  IMO it's a bug.

Wolfram> OK, I see your point now, thanks.  If setting MALLOC_MMAP_MAX_<=0
Wolfram> (whether via environment or via mallopt) would mean _no mmap
Wolfram> whatsoever_, would that be enough (IIRC that's the way used in Emacs,
Wolfram> for example)?  I would submit a patch quickly then.  I'll update the
Wolfram> documentation as well.

It would be according to the docu:
@item M_MMAP_MAX
The maximum number of chunks to allocate with @code{mmap}.  Setting this
to zero disables all use of @code{mmap}.
@end table

Please send a patch,

Thanks,
Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.rhein-neckar.de