Re: tempMem implementation was: Re: "Rude" behavior by developers

New Message Reply Date view Thread view Subject view Author view

From: Moritz Borgmann (moriborg@EIKON.TUM.DE)
Date: 13 Oct 98, 01:40 EST


From: Moritz Borgmann <moriborg@EIKON.TUM.DE>
Subject: Re: tempMem implementation was: Re: "Rude" behavior by developers

> (2)  But the other, more delicate crash, with tempMem ON, is not "raised"
> by you, explicitly enough, in my opinion.  You wrote:
[snip]
> For a programmer (I am speaking in general) to be able to fix a particular
> problem, he absolutely needs to reproduce it on his machine.  (Generally
> speaking, that is.  There are, of course, exceptions.)

I think that this is an exception. As I have pointed out before, quite a
likely cause is simply that the minimal largest free contiguos block in the
application heap is far too small. Because this situation is just
dangerous, the crashs happen very eratically - the bug is not reproducible
by definition.


Masatsugu Nagata wrote:
> On _my_ machine, even when freeMem is very small (like 20K), I open another
> hundred-kilobytes text file and the tempMem is working correctly.  Even
> if I don't open any new file, but continue working on a file, I notice
> freeeMem suddenly grow from 20K to 500K, when I run some Alpha proc

I have not seen such a behaviour yet. Even if Alpha sometimes behaves like
you described - I can perfectly crash Alpha by copying and pasting chunks
of text to fill up the heap. When I paste just small chunks of text, Alpha
still allocates memory in the application heap although memory is already
pretty low. I can reproducibly provoke situations with some 3 KB free. This
itself is _clearly_ a bug, I don't see any need to investigate further.

Moritz

_____________________________________________
Moritz Borgmann
moriborg@eikon.tum.de
http://www.eikon.tum.de/~moriborg/


New Message Reply Date view Thread view Subject view Author view

This archive was generated by das@users.sourceforge.net with hypermail 2b29 on 31 Mar 01, 11:50 EST