Reasonable enough. I've looked into DAFFODIL-2354 and it doesn't look like it is an easy to fix. It appears to be caused by marks that aren't directly related to a point of uncertainty and discarding/resetting newVariableInstance changes. I'm not sure how often that happens in our code base, but there's at least one in choice dispatch which causes this problem.
PR#387 has been merged. We still need one more +1 for PR#389. Considering that is a licensing issue I think it's worth getting in. Anyone have any thoughts on PR#386? I had some comments in the review wondering if an SDW is the right thing for a recoverable error or if we need a special "RecoverableError" construct. Josh is on vacation this week so if we do want that change it likely won't get fixed until at least next week, which feels like an unnecessary delay. Perhaps we just wait for PR#389 and call it? Would anyone like to be the release manager? I don't mind doing it if no one else has the time right now. - Steve On 6/22/20 8:33 PM, Beckerle, Mike wrote: > I agree we can/should create a 2.7.0 for smooks and others. > > If the newVariableInstance bugs are easy, then fixing them is fine, but > otherwise I would suggest just releasing without those fixed. People have > been living without newVariableInstance for a long time. > > > > > > ________________________________ > From: Steve Lawrence <slawre...@apache.org> > Sent: Friday, June 19, 2020 10:38 AM > To: dev@daffodil.apache.org <dev@daffodil.apache.org> > Subject: [DISCUSS] Release Daffodil 2.7.0 > > The smooks project (https://smooks.org) is planning a release shortly > that incorporates Daffodil, but is currently blocked on > DAFFODIL-2359/PR#387. It's been about 2 months since the last release, > and a few useful features and bug fixes have been made, so I wanted to > start a discussion about doing an interim release to the big 3.0.0 that > is planned but is probably still a ways out. > > I suggest that we drop the current Daffodil version in build.sbt to > 2.7.0 and create a new release and move all resolved issues to a 2.7.0 > fix version. Any thoughts/objections? > > > There are currently three open pull requests that I think are close to > being merged and are worth waiting for: > > [PR#386] Implemented dfdl:assert failureType="recoverableError" > > https://github.com/apache/incubator-daffodil/pull/387 > > [PR#387] fixed overflow ArithmeticException when totalDigits >= 10 > > https://github.com/apache/incubator-daffodil/pull/387 > > [PR#389] Update license information for Scala dependencies to Apache v2 > > https://github.com/apache/incubator-daffodil/pull/389 > > > I also think we should at least investigate these two issues: > > [DAFFODIL-2354] newVariableInstance leads to mark state exception > > https://issues.apache.org/jira/browse/DAFFODIL-2354 > > [DAFFODIL-2352] dfdl:newVariableInstance with non-constant defaultValue > fails > > https://issues.apache.org/jira/browse/DAFFODIL-2352 > > > The first issue should almost certainly be fixed since there is no known > workaround. The second would be nice to be fixed, but has a workaround > of using setVariable so maybe isn't quite as important. > > Thoughts on an interim release, and are there any other open issues that > should be resolved for such a release? > > Also, does anyone want to volunteer to be a release manager? Lola did it > last time and did a great job at plugging gaps in the release workflow > documentation, so it should now be a fairly painless process. >
