Proposed patch: smbtorture4: RAW-OPLOCK's CHECK_VAL

Volker Lendecke Volker.Lendecke at SerNet.DE
Wed Jul 25 08:21:22 GMT 2007


On Tue, Jul 24, 2007 at 02:29:39PM -0700, Zack Kirsch wrote:
> I've been playing around with smbtorture and have noticed a bit of an
> oddity in the RAW-OPLOCK test suite.  Specifically, if you have two
> errors in a test (one of them caught by CHECK_VAL), the error from
> CHECK_VAL will be overwritten! I propose the following patch, where we
> will still continue the rest of the test (after failing a CHECK_VAL) but
> will immediately print out the error.

Setting "kernel oplocks = yes" in make test on Linux:
Without your patch I for example get

TEST BATCH10 FAILED! - torture/raw/oplock.c:964: wrong value for break_info.fnum got 0x0 - should be 0x106c

With your patch: 

(torture/raw/oplock.c:963): wrong value for break_info.count got 0x0 - should be 0x1
(torture/raw/oplock.c:964): wrong value for break_info.fnum got 0x0 - should be 0x109e
ERROR IN TEST BATCH10! - Unknown error/failure

I don't know how to solve this Unknown error/failure thing.
The "real" error message should have been "Hey, you did not
send an oplock break".

Not applying directly, I'm not sure which alternative is the
better one.

Volker
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.samba.org/archive/samba-technical/attachments/20070725/37acb70f/attachment.bin


More information about the samba-technical mailing list