Spec says that writer should be closed before reader. RI does things in reverse.

[follis]

Originally opened as bug 6456 by ScottKurz

--------------Original Comment History----------------------------
Comment from = ScottKurz on 2014-10-07 11:02:03 +0000

---

Comment from = ScottKurz on 2014-10-23 16:27:23 +0000

Oh..one more minor note:

In "Rollback Procedure" at the end of Sec 11.9 we for some reason have the writer open() call before the reader open().

This seems like a copy/paste error. I would like to queue up a spec update to treat it as such.

Any objections?

---

Comment from = ScottKurz on 2014-10-30 20:59:04 +0000

Fixed in RI (along with other fixes) at:
WASdev/standards.jsr352.jbatch@1679181

Moving this to 'SPEC' to address the typo in Comment 1.

---

Comment from = struberg on 2015-09-07 07:52:26 +0000

was there a test for it in the TCK or was it described that way in the API?
If so, then we need to keep it imo (according to JCP rules).

If it was ambiguous or even contradictory then we are free to 'fix' it.

---

Comment from = BrentDouglas on 2015-09-07 11:14:44 +0000

There is no TCK test for it, or if there is, it is broken.

I don't know if it's ambiguous, section 11.9 says has ItemWriter.open before ItemReader.open, but it does go against the convention every other time these interfaces are mentioned where the pattern is always ItemReader.open, ItemWriter.open, ItemWriter.close, ItemReader.close.

---

Comment from = ScottKurz on 2016-03-10 22:15:07 +0000

(In reply to struberg from comment #3)

was there a test for it in the TCK or was it described that way in the API?
If so, then we need to keep it imo (according to JCP rules).

If it was ambiguous or even contradictory then we are free to 'fix' it.

No, I'm almost 100% certain there was no test specifically enforcing the relative order of reader vs. writer open() in "Rollback Procedure". (As there was no test enforcing the order during normal processing, given that the RI contained this bug in 1.0).

So I think we're free to update the spec doc. And we really should add some TCK test coverage too.

(This order still surprises me though.)