crash recipe for htmlText?

[xbury.cs at clearstream.com]

Just in case, you wouldn't have a line extending past the safe line-size 
limit?
I've seen some strange html sources cause problems previously because of 
this.

Good luck anyway...

regards,
Xavier

On 11.01.2005 09:04:54 use-revolution-bounces wrote:
>xbury.cs at clearstream.com wrote:
>> On 10.01.2005 22:58:57 use-revolution-bounces wrote:
>>
>>> I'm getting some intermittent hard crashes after setting
>>> the htmlText of a field.  Trying to pin this down, I have
>>> a script that repeatedly performs the same parsing,
>>> formatting, and htmlText setting over and over, and
>>> sometimes all of it works well, sometimes it displays
>>> garbage and crashes, and sometimes it displas garbage
>>> and doesn't crash.
>>
>> Some incompletely loaded images used to create crashes.
>> But haven't seen it since 2.5.
>>
>> Other than that, and with extensive testing using the
>> DiscreteBrowser I haven't had a problem with anything.
>>
>> Care to send me the html to test it here?
>
>That's the hard part:  there isn't one.  It's a large collection, and
>sometimes it works great for several thousand iterations, then other
>times I'll see garbage in a field without a crash, and sometimes garbage
>with a crash.  It crashes on different files, and after a different
>number of iterations each time. :\
>
>I think the thing about images may get us closer to understanding this
>anomaly.  In my own tests any gabage that appears is always after the
>first image referenced in my htmlText.  I've been running a test in the
>background in which I'm not including in the image references, many tens
>of thousands of iterations without a crash.  Hmmm....
>
>It may make a difference with the crash (and certainly with subsequent
>performance) to redesign the main routine to cache the files and
>reference the local cached copy for display.
>
>I needed to add that anyway, so I'll do that now and report back on how
>it works out.
>
>On the upside I know this can be done well because WebMerge has been a
>similar method for years, and with many copies out in weekly use (with
>some customers daily) it's never crashed yet.
>
>Hopefully I'll have good news and a recipe or workaround to report by
>morning...