On 30/03/2010, Konstantin Kolinko <knst.koli...@gmail.com> wrote:
> 2010/3/30 sebb <seb...@gmail.com>:
>
> > On 29/03/2010, ma...@apache.org <ma...@apache.org> wrote:
>  >> Author: markt
>  >>  Date: Mon Mar 29 15:01:14 2010
>  >>  New Revision: 928798
>  >>
>  >>  URL: http://svn.apache.org/viewvc?rev=928798&view=rev
>  >>  Log:
>  >>  Fix https://issues.apache.org/bugzilla/show_bug.cgi?id=48895
>  >>  Make clearing thread locals optional and disabled by default since it 
> isn't thread-safe
>  >
>  > However, the new boolean flag (clearReferencesThreadLocals) that
>  > controls the process is not safely published between threads. It
>  > should probably be volatile.
>  >
>
>
> Startup is performed by a single thread, and it will set the value of
>  this field.

OK, in that case not a problem.

However, it seems to me that this assumption should be documented in
the StandardContext class.

I think I saw a recent Bugzilla which suggested using multiple threads
to speed startup, in which case the assumption might become incorrect.

>  According to the memory model all the threads started after that will
>  see the assigned value.

Agreed; it's only a problem if two active threads access the same
mutable variable.

>  I do not remember any of the "option value" fields in Tomcat being
>  marked as volatile. All of them are usual fields.
>
>  Best regards,
>
> Konstantin Kolinko
>
>
>  ---------------------------------------------------------------------
>  To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
>  For additional commands, e-mail: dev-h...@tomcat.apache.org
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org

Reply via email to