[
https://issues.apache.org/jira/browse/DERBY-3735?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kristian Waagan updated DERBY-3735:
-----------------------------------
Attachment: derby-3735-2a.diff
'derby-3735-2a.diff' fixes the bug introduced by the previous patch (1b)...
0 (zero) is of course a valid return value for "read()", and the position has
to be incremented.
I changed the "read(byte[],int,int)" to check for -1 for consistency as well.
This method will return 0 very seldom, in fact I think it requires a "user
error" to happen; argument 'len' must be zero.
Regression tests run without failures.
Patch ready for review, but I will commit it shortly anyway to get rid of the
tinderbox/nightly failures.
> Incorrect position calculation in PositionedStoreStream with read(byte[],...)
> -----------------------------------------------------------------------------
>
> Key: DERBY-3735
> URL: https://issues.apache.org/jira/browse/DERBY-3735
> Project: Derby
> Issue Type: Bug
> Components: JDBC, Store
> Affects Versions: 10.3.1.4, 10.4.1.3, 10.5.0.0
> Reporter: Kristian Waagan
> Assignee: Kristian Waagan
> Priority: Minor
> Fix For: 10.5.0.0
>
> Attachments: derby-3735-1a.diff, derby-3735-1a.stat,
> derby-3735-1b.diff, derby-3735-2a.diff
>
>
> A bug in the methods 'read(byte[])' and 'read(byte[],int,int)' in
> PositionedStoreStream can cause the position variable to be set to an
> incorrect value.
> The bug is only triggered if one of the two read methods is invoked after EOF
> of the underlying stream has been reached.
> If the bug is triggered, the position will be decreased by one because the
> underlying stream returns -1. Subsequent reads will further decrease the
> position. This can cause EOF-exceptions (during repositioning) or incorrect
> data being returned in subsequent calls.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
