Andrew McIntyre wrote:
On 6/20/06, Kathey Marsden <[EMAIL PROTECTED]> wrote:
Mike Matrigali wrote:
> I would like to see the fix for DERBY-1392 included in the 10.1.3
> release if there is a second release candidate. While the bug
> is an edge error case, the result is a corrupt db. I believe
> there is little risk as again the path is not one usually taken.
> The change has already been fixed in the trunk and the 10.1 branch.
+1 to including DERBY-1392 and thanks so much to Anders for finding and
fixing this issue!
Is this a serious enough issue to warrant another release candidate?
Tests that exercise the issue were not contributed along with the fix,
and it would be nice to know that this is an issue that is likely to
be hit in known circumstances. If so, a release note is probably in
order since it supposedly can cause corruption to a database.
I believe this is serious enough to warrent another RC. By code
inspection as Anders points out it is clear there is a path that
could corrupt the db.
Another reason I would like to get this in is that in the past there
has been a non reproducible case of an unencrypted string getting
into a page of an encrypted db with one of our customers on the
original cloudscape pre-derby product. We have
never been able to reproduce this, but the code for this path is
has the same problem in the pre-derby code as the code Anders fixed.
Since we never had a reproducible case I can't say this is the fix,
but it seems likely.
Because the code path depends on a timing I/O exception it is a
particularly hard test to write. For this reason I spent extra
time reviewing the code and why I felt it was ok to go in without
a test.
Since 10.2 is around the corner, I am worried that this might be the
last official 10.1 release for a long time so would like to see this
fix make it.
