[
https://issues.apache.org/jira/browse/DAFFODIL-2883?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mike Beckerle resolved DAFFODIL-2883.
-------------------------------------
Resolution: Fixed
Fixed in *8433f4d83d33da31371fcdfa016a7c627011aca3*
There is no switch added. The fix is just to make the conversion as called by
the InfosetOutputter, always accept pre-existing PUA characters.
The DFDL Infoset allows PUA characters, so the InfosetOutputter should
definitely *not* fail on them ever.
> Pre-existing PUA characters in data cause SDE
> ---------------------------------------------
>
> Key: DAFFODIL-2883
> URL: https://issues.apache.org/jira/browse/DAFFODIL-2883
> Project: Daffodil
> Issue Type: Bug
> Components: Back End
> Affects Versions: 3.6.0
> Reporter: Mike Beckerle
> Assignee: Mike Beckerle
> Priority: Major
> Fix For: 3.8.0
>
>
> If data contains Unicode PUA characters, these cause the Infoset Outputter to
> convert the RemapPUACharDetected into an SDE.
> We can't get away with this. We need to tolerate PUA characters in data and
> have them either cause a ParseError, or just tolerate them. (Or have a switch
> to choose between those modes)
> This was discovered by fuzz testing.
> If the existence of PUA characters means the data is gibberish, then perhaps
> the parser is speculating down a path that should be backtracked. We need a
> parse error in that case.
> If the existence of PUA characters is acceptable, then we need no error at
> all from them.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
