API changes in patch release

Thomas Hallgren wrote:

The world is not perfect and I know that you are normally very
restrictive in what is back-patched from head into bug-fix branches. The
8.1.4 release however, did introduce a problem. You changed the API
function inv_open() with the comment "Revise large-object access
routines to avoid running with CurrentMemoryContext".

This change will force me to a) introduce patch level sensitive
conditionals in the code, and b) have two PostgreSQL 8.1.n compatible
releases of PL/Java. One where n < 4 and another where n >= 4. I would
like to avoid this in the future if possible. API's should remain stable
during patch releases.

Having said that, I've been in the game long enough to know that Utopia
doesn't exist. You probably had a very good reason to break the
compatibility.

I don't think any of us realized the change would affect third-party
projects.

--
Bruce Momjian http://candle.pha.pa.us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

This change will force me to a) introduce patch level sensitive
conditionals in the code, and b) have two PostgreSQL 8.1.n

compatible

releases of PL/Java. One where n < 4 and another where n >=

4. I would

like to avoid this in the future if possible. API's should remain
stable during patch releases.

Having said that, I've been in the game long enough to know that
Utopia doesn't exist. You probably had a very good reason

to break the

compatibility.

I don't think any of us realized the change would affect
third-party projects.

To help specifically PL/Java next time, is there any chance to get it
included in the buildfarm builds? If it had been there, it would've been
caught right away...

//Magnus

Magnus Hagander wrote:

I don't think any of us realized the change would affect
third-party projects.

To help specifically PL/Java next time, is there any chance to get it
included in the buildfarm builds? If it had been there, it would've been
caught right away...

Currently buildfarm builds the core, and nothing else. Testing of non
core code (pgfoundry projects, uncommitted patches, unknown other stuff)
is on a possible TODO list. It's something I want discussed at least
informally at the Toronto conference - I hope the organizers have
managed to find me a decent slot for a small discussion session to cover
buildfarm issues.

Right now I do not have enough time available to do much development
work on it, though.

cheers

andrew