It potentially changes every minor release (in particular the internal interfaces like
SessionImplementor that we need to implement).
One of the benefits is that an old version of HSearch would work with the latest version
of Hibernate Core including the new methods and we would not have to maintain an complex
matrix of working versions.
But I agree it's mostly a development problem and I happen to be the one hitting it
all the time :)
On 13 janv. 2010, at 16:21, Hardy Ferentschik wrote:
Actually I've never had this problem. Admittedly after reading
your email I did a quick
check and I get also a compilation error due to 'isReadOnly' not being
overridden.
Does Session really change that often that we have a use of a dynamic proxy? Would it
not be easier to not depend on a SNAPSHOT in the pom. We could depend on the latest 3.5
beta. You still have to implement the missing methods in FullTextSession when you do
an upgrade, but the process is more controlled.
Are there any other benefits of a dynamic proxy?
--Hardy
On Wed, 13 Jan 2010 12:07:48 -0300, Emmanuel Bernard <emmanuel(a)hibernate.org>
wrote:
> Guys,
> I am sick of HSearch being broken every time Core adds a new method to Session /
SessionImplementor and the like.
> I am seriously thinking about providing a dynamic proxy implementing FTSession etc
and will try and prototype it.
>
> Any opinion?