OK, I made all my tests with 2.1b10 again.
No crash and no problems with my japanese text arrays in this command anymore.
I forgot which version is was, but I think 2.1 b3-6.
It works, and forget this problem.
Now that has got me thinking in what the differences between the previous
caching code and the crashing caching code are.
The code that rarely crashes will probably not include characters over ASCII
128, or is real or Longint arrays. I say probably because what was stored
was the query parameters and I *believe* that I store the unconverted
version.
The data that was guarenteed to crash Active4D will be very likely to
contain characters with accents, or other non printable low ASCII value
characters. Since this was product names and descriptions in German. So
probably items with >128 ASCII and possibly lower than ASCII 32 due to the
source of this data.
Ken Ishimoto
