Re: odd awk memory leak?

[Mouse <mouse%Rodents-Montreal.ORG@localhost>]

>>> That's an effect of the changed malloc implementation that no
>>> longer takes memory from the 'data segment'.
>> That strikes me as a bit of a bug, or at least a regression, if the
>> data size limit is no longer an effective way to bound the memory
>> used by a naïve program that just malloc()s.
> There is a new address space limit in place.

"vmemoryuse"?  That helps, but it's still a nasty surprise lurking for
anyone used to using data size limit to, well, limit data size.
Especially since vmemoryuse defaults to unlimited (at least on my 5.2
machine it does).

> I'd say it is just as surprising that malloc changed the data
> segment...

In isolation from history, I might agree.  But there is a lot of
history behind data size limits limiting, well, data size.  Whether
that data is collected into a single "segment" or not is, it seems to
me, a historical and/or implementation artifact.

It's also logically coherent to think of the datasize limit as being a
limit on just the data segment part of that; that seems to be what's
behind this - but that seems to me to be a less useful point of view,
in that the implementation that proceeds from it has more surprises
waiting for people.  To preserve behaviour across the version change
requires user-, or at least programmer-, -visible changes which are
compatible in neither direction.

/~\ The ASCII                             Mouse
\ / Ribbon Campaign
 X  Against HTML                mouse%rodents-montreal.org@localhost
/ \ Email!           7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B