Guillaume, it depends unfortunately. In distributed cases, checking the
status of a transaction could mean remote calls. That's why I was saying
I'd rather not have the unnecessary overhead if not needed.
On Thu, Oct 30, 2014 at 9:51 AM, Guillaume Smet <guillaume.smet(a)gmail.com>
wrote:
On Thu, Oct 30, 2014 at 2:56 PM, Steve Ebersole
<steve(a)hibernate.org>
wrote:
> Personally having entities dirtied as part of a read-only transaction
sounds
> like an application bug to me. We could try to detect a read-only
> transaction state (not sure how we'd do that across all cases) and
> circumvent the flush there, but that would add unnecessary overhead to
> applications that do the right thing.
They aren't dirtied as part of the read-only transaction per se but I
agree with you it's more an application bug (it's due to the way we're
dealing with empty embedded).
Just wanted to report it in case it was easy and free to test the
transaction status!