Update on Safeguard

["John D. Ament" <jo...@apache.org>]

Hey guys

I pushed up a pretty big commit last night which adds bulkhead support.
Lots of pain in the TCK in that area, if it was clearer what was going on I
would have been done a few weeks ago :-(

Anyways, with that commit the last actual Fault Tolerance feature is done.
There's two areas left:

- Configuration overriding
- Configuration validation on deployment

I already know there's an issue with the TCK in the validation area.  I
still plan to add it, but the TCK tests likely won't pass due to an
Arquillian bug (the issue I previously mentioned on list).

With configuration overriding there's a single TCK test that only checks
two attributes are configurable.  The spec makes a blanket statement.  I'm
going to create additional tests to verify everything.  However, there's
one implicit issue.

The spec, with the blanket statement, is assuming that I can convert
probably from a comma separated String into a Class<? extends Throwable>[]
as well as convert to a ChronoUnit.  I see two approaches.  One is to put
those converters into GConfig the other is to put them into Safeguard.

I'm of the opinion that if FT needed these converters, they should have
ensured they were built in to config.  This would have ensured that proper
formats were supported.

John

Re: Update on Safeguard

[Mark Struberg <st...@yahoo.de>]

If we can push this change to mp-config and JSR-382 then I'd put it into config.
LieGrue,strub


    On Saturday, 28 October 2017, 14:09:37 GMT+2, John D. Ament <jo...@apache.org> wrote:  
 
 Hey guys
I pushed up a pretty big commit last night which adds bulkhead support.  Lots of pain in the TCK in that area, if it was clearer what was going on I would have been done a few weeks ago :-(
Anyways, with that commit the last actual Fault Tolerance feature is done.  There's two areas left:
- Configuration overriding- Configuration validation on deployment
I already know there's an issue with the TCK in the validation area.  I still plan to add it, but the TCK tests likely won't pass due to an Arquillian bug (the issue I previously mentioned on list).
With configuration overriding there's a single TCK test that only checks two attributes are configurable.  The spec makes a blanket statement.  I'm going to create additional tests to verify everything.  However, there's one implicit issue.
The spec, with the blanket statement, is assuming that I can convert probably from a comma separated String into a Class<? extends Throwable>[] as well as convert to a ChronoUnit.  I see two approaches.  One is to put those converters into GConfig the other is to put them into Safeguard.  
I'm of the opinion that if FT needed these converters, they should have ensured they were built in to config.  This would have ensured that proper formats were supported.
John