[arachne] Re: [a4dos-list] Internal Editor.. more musings

["Greg Mayman" <gmone@xxxxxxxxxx>]

Arachne at FreeLists---The Arachne Fan Club!

On Thu, 27 May 2004 10:31:29 +0100, John Sparks wrote:

> Arachne at FreeLists---The Arachne Fan Club!

> My theory (feel free to rubbish if ill conceived):

> If there is a bug in Arachne's memory handling the same area might
> under some circumstances be allocated to two different needs. If the
> editors work space was double booked then that data would be
> vulnerable to ongoing corruption. A subsequent editor session might
> not be double booked and would remain unaffected.

Yes, that makes sense. I was thinking more or less along that line too.
And since it only plays up about once in every 10-20 times, I shall try
to live with it.

> Different users will have different quantities of memory available
> which could affect the pattern of memory allocation. This would be
> consistent with different reported experiences.

Yup, I'm with you so far.

> When an external editor was called Arachne did some work preparing to
> write the already corrupted data to a file for passing to the external
> editor. This could result in further corruption before the file is
> written to disk.

Yes, I have noticed that effect.

> The corruption would be bi-directional. When double booked, Editor
> input could trample upon other data. This might? explain the reported
> problems relating to the file name used for external editor data.

Yes, although this was the one and only time I have seen that
happen.

> Difficult to track down without some hard data. Would it be possible
> to build a debug version of Arachne in which a selective memory dump
> could be invoked by the user? This might allow a developer to view how
> the memory was allocated when the problem was present, and obtain some
> clues.

I'm not sure that a memory dump would be needed, or useful. 

There would be such an amount of stuff to wade through it would
be difficult to find where the trouble had occurred.

Perhaps the memory allocation would be better. It could be
automatically written continuously every time anything was
written to memory.

But since anything like this would require the use of memory, it
becomes another case of the very fact of the observation altering
the subject of the observation.

The correct way of course is to go through Arachne's code with a
fine-toothed comb, and find the place that causes this memory
conflict.

And be assured that it ain't gunna be at all obvious. It would
almost certainly be quicker and easier to rewrite all of Arachne's
code from scratch.... v2.0....

   ,-./\      from Greg Mayman, in Adelaide, South Australia
  /     \     "Queen City of The South"    34:55 S  138:36 E
  \_,-*_/
       v     homepage http://users.tpg.com.au/gmone/index.htm


Arachne at FreeLists
-- Arachne, The Web Browser/Suite for DOS and Linux --

• Follow-Ups:
  • [arachne] Re: Internal Editor.. more musings
    • From: John Sparks

Other related posts:

[arachne] Re: [a4dos-list] Internal Editor.. more musings

[arachne] Re: [a4dos-list] Internal Editor.. more musings