Friday, March 31, 2006

There are bugs and BUGS

Dmitry Yemanov writes:

Consider it a bughunting success story :-) Some time ago we've got a bugreport about bugcheck #205 (index bucket overfilled) appearing during the database restore on a v2.0 server. The database was big and it was hard for the customer to create a test case. Fortunately, a special build with improved tracking facilities helped and the bug has been found and fixed. So far so good. But a few weeks later, this bug reappeared on the same database and under the same conditions. The only difference was a database size which has grown. After unsuccessful attempts to find the bug, the reduced copy of the customer database (with just a single table that causes the problem) has been received. It reproduced the bug successfully, with just one minor issue... it was a 3.5GB backup file :-) The restore process takes 6 hours in the debugger. It required a few runs to figure out the actual problem (don't ask me how many days it has taken :-), but it has been finally fixed. Sigh. Some bugs are buggier than others :-) By the way, this should be exactly the bug that prevented Dalton Calford from restoring his database on v2.0 (he reported the issue in fb-devel).

No comments: