[
https://issues.apache.org/jira/browse/DAFFODIL-1595?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mike Beckerle reassigned DAFFODIL-1595:
---------------------------------------
Assignee: Mike Beckerle
> Check for data too long for specified length - missing
> ------------------------------------------------------
>
> Key: DAFFODIL-1595
> URL: https://issues.apache.org/jira/browse/DAFFODIL-1595
> Project: Daffodil
> Issue Type: Bug
> Components: Back End, Diagnostics, QA, Unparsing
> Reporter: Mike Beckerle
> Assignee: Mike Beckerle
> Priority: Major
> Labels: reverify
> Fix For: 3.5.0
>
>
> Due to the out-of-order way that dfdl:outputValueCalc has to work, the way
> unparsers compute how long data is, that had to change.
> That results in the means by which the unparsers measure and detect that the
> data is too long, need to be different.
> Now, there used to be code in unparser primitives that did this checking, but
> such code can't work in general anymore.
> The place that has to do this checking now is centralized into the
> RightFill(simple types) and ElementUnused(complex type) unparsers. However,
> this check isn't implemented there.
> No tests fail, so either there aren't tests exercising this, or the ones that
> are exercising this just happen to run because they do not also exercise
> dfdl:outputValueCalc, so nothing out-of-order is happening, and so the older
> checking code is maybe still working in those specific cases?
> So two things (a) checking code needs to be added to RightFillUnparser and
> ElementUnusedUnparser to detect when data that cannot be truncated is being
> unparsed into a specified length that is too small for it (b) tests that
> create errors due to data too long need to be created which also use aspects
> of the unparser that force the DirectOrBufferedDatatOutputStream to be used
> underneath, so as to force this error to be detected.
> Part of (a) is removing the legacy checking code which is not general enough
> to work now, from the few unparser primitives where it appears (might be only
> one?)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
