http://ayende.comAyende @ RahienCopyright (C) Ayende Rahien 2004 - 2021 (c) 202660The code isn't the best, but the error is real http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment12http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment12Fri, 08 Aug 2008 15:18:40 GMTCould it be a bug in your code? If I understand correctly, in the exception catch you only test for one message, for other messages control falls through to done = true and ends the loop, returning the data that was read in the first select command. http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment11http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment11Thu, 07 Aug 2008 22:45:45 GMTI'll let you guess what I am working on... http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment10http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment10Wed, 09 Jul 2008 03:56:36 GMTThis comment: // someone else already grabbed and deleted this row, // so we will try again with another one is shouting semaphore. Even then, I'm hearing MSMQ whispering in the wind. http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment9http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment9Tue, 08 Jul 2008 19:56:34 GMTYou have the SQL there, and the table structure is 1:1 with that. Make sure to run this on 3 threads, that is how I got it. http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment8http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment8Tue, 08 Jul 2008 15:21:01 GMTSerializable and ReadCommited are both supported in FireBird. ReadCommited is default isolation. http://www.firebirdsql.org/dotnetfirebird/blog/2005/02/transaction-isolation-levels-in.html could you provide some really simple set of SQL/DDL commands to simulate the error behavior you describe? http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment7http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment7Tue, 08 Jul 2008 12:07:45 GMTThere isn't a second select, it is a delete. Basically, I am doing: 1) give me the first row 2) delete the previously selected row 3) if successfully deleted, return row 4) if failed to delete, go to 1) That is safe in read committed mode, but it fails with FireBird. http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment6http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment6Mon, 07 Jul 2008 11:57:20 GMTYou mean, the second select shouldn't succeed if someone else already read the row which should have placed a read lock on the row? I'm not sure if Firebird sets read-locks in such a way that multiple reads of the same row isnt possible if the row was already locked for reads. I'm also not sure if firebird supports hints for locks. Their manual is pretty bad, to say the least. http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment5http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment5Mon, 07 Jul 2008 11:50:54 GMTFrans, The code above should work with even with read committed. Notice that I am explicitly checking that the row was removed, and only then moving on. I still managed to get duplicate results here. 1) I often use this approach to handle additional concerns. Consider this as a poor man's AOP. 2) I disagree with that. All access to the DB should be done under a transaction. http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment4http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment4Mon, 07 Jul 2008 07:51:56 GMTDuh, firebird doesn't have that isolation mode (only some databases do). What I DO find disturbing in your code are: 1) a big chunk of code in a lambda. IMHO it's better to write a separate routine for that but alas 2) a fetch inside a transaction. Fetches inside transactions are a clue that something is out of order: fetch first, then start transaction. http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment3http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment3Mon, 07 Jul 2008 07:48:46 GMTThat is not the issue, the issue is that there is no transaction isolation maintained http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment2http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment2Mon, 07 Jul 2008 07:25:48 GMTTry e.Number or e.Errors[0].Number See: http://www.firebirdsql.org/dotnetfirebird/documentation/api/1.7/FirebirdSql.Data.Firebird.FbException.html http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment1http://ayende.com/3407/why-i-dont-like-firebird-part-ii#comment1Mon, 07 Jul 2008 06:19:14 GMT