Bryan Pendleton wrote:
Sunitha Kambhampati wrote:
they are hitting an edge case, where the writer.getDSSLength() is
equal to blksize and this DSS is the only dss in the send buffer.
Below is a simple repro program that you can run to demonstrate the
bug for yourself. Compile it, then start up the Network Server and
run the program with "java newBug 12749". That exact length for column
"c2" will trigger the edge case you've found.
Thanks very much Bryan for jumping on this and providing the repro.
Thats just great!
The fix is as simple as changing the >= to >, just as you said in your
message, I believe.
Yes. Thats what I tried and it solves the user's problem. I also ran
derbynetmats and derbynetclientmats and all tests passed.
Should we re-open one of the existing bugs? I think it may be easier just
to file a new bug, and then verify the fix using the diff and the test
program below.
I am fine either ways. If opening a new jira is easy, then we should
just do that. Since DERBY-170 is already backported to 10.1, we will
need to backport this fix to 10.1 also. If we go with opening a new
issue, it might be good to link this new issue with derby-170 using jira
links, so that in the future we have this history available for us.
I'll work this up into a proper patch, with a new regression test,
probably as part of derbynet/prepStmt.java
Excellent. Thanks again.
please let me know if you can confirm these results.
Yes, I can repro this on my machine. The server hits this offbyone bug
and disconnects the client.
Thanks,
Sunitha.
