Re: concurrent access deadlock

Philip Martin <pmartin@uklinux.net> writes:

> The deadlock I reported in
>
> http://subversion.tigris.org/servlets/ReadMsg?msgId=42676&listName=dev
>
> where two separate working copies commit to the same repository, seems
> to be easy to reproduce at present. I am building against db-4.0.14 as
> an experiment, and the script in the mail above now provokes deadlock
> fairly often. I don't think 4.0.14 is itself the cause of the deadlock
> as I could sometimes provoke it when using 3.3.11 (but I may be wrong
> here :-)

Boy, it's been a while since I read this mail from you. Has any new
progress been made? Is it still easy to reproduce?

> I tried running 'db_recover -h repostress/db' but that appears to
> destroy the database, as a subsequent 'db_stat' produces
> db_stat: DB_ENV->open: repostress/db: No such file or directory

Just so you know, when using db_recover, it's best to use it with:

   db_recover -ve -h <db_path>

That way, it rebuilds the environment (and doesn't "destroy the
database"). Ben helped me to remember those extra arguments -- around
here we ask, "Did you run db_recover-dash-very-exciting?"

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:56 2006