Re: Thread-safe messages
Thanks Cristopher and Kenton! I will provide a Mutex to the higher levels as you recommend. Jes. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Protocol Buffers" group. To post to this group, send email to protobuf@googlegroups.com To unsubscribe from this group, send email to protobuf+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/protobuf?hl=en -~--~~~~--~~--~--~---
Re: Thread-safe messages
Simply adding a mutex lock in every accessor wouldn't really make them
thread-safe. Consider:
if (my_message.has_foo()) {
DoSomething(my_message.foo());
}
This is not thread-safe if my_message can be modified in a separate thread
*even if* each accessor locked a mutex, because "foo" could be cleared in
another thread between the call to has_foo() and the call to foo().
Here's a more dangerous example:
for (int i = 0; i < my_message.foo_size(); i++) {
DoSomething(my_message.foo(i);
}
Here, if "foo" is cleared between the call to foo_size() and the call to
foo(i), then the program will crash. Having a mutex protecting each
individual accessor does not help.
Another problem is that strings are returned by reference. So even if the
accessor itself locks a mutex, the string's contents may be accessed after
the accessor returns, and if they are simultaneously modified in another
thread you'll get a crash.
What you really need to do is maintain your own mutex at a higher level.
Any thread which reads the message object needs to take out a reader lock
before it starts reading and hold that lock until it is completely done
accessing the object. Similarly, any thread writing to a message object
needs to take a writer lock on that mutex at a higher level.
The protobuf library can't really do anything to help here, since the
correct mutex usage depends entirely on your application code.
On Mon, Jun 29, 2009 at 10:01 AM, Jes wrote:
>
> Hi everybody,
>
> we are working on a distributed environment that uses PB, where
> different threads will access to the contents of messages that can be
> updated at any moment through the network.
>
> I wonder if there is an easy way to transform the (derived) Messages
> into a thread-safe version. Maybe the rough solution could be to
> include a Mutex in the Message class and a MutexLock on each method of
> the generated pb.h and pb.cc classes, but perhaps there are issues
> that can break the safety of this approach (such as existing friends
> or similar).
>
> Could you have any suggestion on this? :-)
>
> Thanks in advance!
>
> Jes
>
>
> >
>
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups
"Protocol Buffers" group.
To post to this group, send email to protobuf@googlegroups.com
To unsubscribe from this group, send email to
protobuf+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/protobuf?hl=en
-~--~~~~--~~--~--~---
Re: Thread-safe messages
I'd recommend using an atomic swap to do your updates. So you create your new version of the PB localy, and then swap it in to the memory location that is visible to all the other threads. The only real downside is you stress the heap more, and that is probably cheaper/simpler (particularly if you want the updates to be transactional) than using extensive locking. --Chris On Mon, 2009-06-29 at 10:07 -0700, Jes wrote: > I forgot to mention that we are generating C++ code in the project. > > Jes > > On 29 jun, 19:01, Jes wrote: > > Hi everybody, > > > > we are working on a distributed environment that uses PB, where > > different threads will access to the contents of messages that can be > > updated at any moment through the network. > > > > I wonder if there is an easy way to transform the (derived) Messages > > into a thread-safe version. Maybe the rough solution could be to > > include a Mutex in the Message class and a MutexLock on each method of > > the generated pb.h and pb.cc classes, but perhaps there are issues > > that can break the safety of this approach (such as existing friends > > or similar). > > > > Could you have any suggestion on this? :-) > > > > Thanks in advance! > > > > Jes > > --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Protocol Buffers" group. To post to this group, send email to protobuf@googlegroups.com To unsubscribe from this group, send email to protobuf+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/protobuf?hl=en -~--~~~~--~~--~--~---
Thread-safe messages
Hi everybody, we are working on a distributed environment that uses PB, where different threads will access to the contents of messages that can be updated at any moment through the network. I wonder if there is an easy way to transform the (derived) Messages into a thread-safe version. Maybe the rough solution could be to include a Mutex in the Message class and a MutexLock on each method of the generated pb.h and pb.cc classes, but perhaps there are issues that can break the safety of this approach (such as existing friends or similar). Could you have any suggestion on this? :-) Thanks in advance! Jes --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Protocol Buffers" group. To post to this group, send email to protobuf@googlegroups.com To unsubscribe from this group, send email to protobuf+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/protobuf?hl=en -~--~~~~--~~--~--~---
Re: Thread-safe messages
I forgot to mention that we are generating C++ code in the project. Jes On 29 jun, 19:01, Jes wrote: > Hi everybody, > > we are working on a distributed environment that uses PB, where > different threads will access to the contents of messages that can be > updated at any moment through the network. > > I wonder if there is an easy way to transform the (derived) Messages > into a thread-safe version. Maybe the rough solution could be to > include a Mutex in the Message class and a MutexLock on each method of > the generated pb.h and pb.cc classes, but perhaps there are issues > that can break the safety of this approach (such as existing friends > or similar). > > Could you have any suggestion on this? :-) > > Thanks in advance! > > Jes --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups "Protocol Buffers" group. To post to this group, send email to protobuf@googlegroups.com To unsubscribe from this group, send email to protobuf+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/protobuf?hl=en -~--~~~~--~~--~--~---
