Re: [Exist-development] Errors from Stress tests on xquery update

[Wolfgang M. <wol...@ex...>]

> Thanks. I will test this workaround. So would you consider this to be a bug
> that will be eventually fixed in eXist, or will developers always need to be
> aware of this when using the update methods?

There seems to be a bug. eXist does allow dirty reads, which means it
is possible for node references to become invalid during a query.
Nevertheless a started transaction must not abort.

I have received another test from someone else which fails for similar
reasons. We have to figure out what causes this. It's not easy though.
If you have a reproducible test, I'll be happy to compare it to the
other scenario I have.

Wolfgang