[elephant-devel] test concurrent

Robert L. Read read at robertlread.net
Fri Apr 11 03:22:43 UTC 2008


Thank you, Alex, that is a very nice solution.  I have committed it to
the main repo.

Yes, I thought it did work in at least one place.




On Thu, 2008-04-10 at 10:51 +0300, Alex Mizrahi wrote:
>  RLR> However, fixing those tests (or the concurrency problem they
>  RLR> represent) under SBCL is a high priority.
> 
> what do you mean by "fixing under SBCL"? do you think they are working on 
> some other implementation?
> very unlikely..
> 
> i've improved and extended concurrency tests, and they work mostly fine with 
> postmodern.
> but still that's quite weird stuff, so i think they should stay separate and 
> optional -- staying in their own suite:
> 
>  (in-suite* testthreads)
> 
> so you can run them explicitly via (do-test-spec 'testthreads)
> 
> tests require serializable isolation level (without it there is no 
> consistency -- indices can be descynchronized from slots and other weird 
> things happen),
> deadlock detection and transactions restarts due to access conflicts and 
> deadlocks. if you want to pass tests with CLSQL, you can start improving it 
> in these areas. but i'm not sure if it's possible at all for all supported 
> backends..
> 
> also, some tests (threaded-object-creation-0-i) fail regardless of 
> backend -- due to index caching we have in elephant.
> i've commented them out for now, but that doesn't mean that issues should 
> not be addressed.
> 
> here's what in attached patch bundle:
> 
> test concurrency extended
> un-disabled tests for SBCL
> 
> db-postmodern: transaction retry handling, serializable isolation
> transaction restart support plus extended features
> 
> db-postmodern: small sync cache fix
> db-postmodern: sync-cache type handling fix
> _______________________________________________
> elephant-devel site list
> elephant-devel at common-lisp.net
> http://common-lisp.net/mailman/listinfo/elephant-devel




More information about the elephant-devel mailing list