Status of the C++ daemon, Moving objects with references to other members of the same class

[Pedro Larroy]

Hi

I spent many hours in a transcontinental flight tracking down a bug on the Share class, was displayed when moving a Share, the sqlite3pp::statements would get moved and the reference to the sqlite3pp::database would be invalidated, as the object to which they refer was destroyed, in this case the database. I identified this pattern two times in this class. This is a C++11 mistake that is not obvious and easy to do. It's fixed in the latest commits to master.

I also got a rudimentary GET command working, still is not pipelined as the protocol suggests.  So I think in the following weeks/month we should be able to use the C++ daemon so synchronize files without encryption.

It would be great if somebody finds the time to improve the status of the cryptographic parts.

I'm excited to getting closer to a first prototype.

Regards.

[Steven Jewel]

On 04/21/2014 01:58 AM, Pedro Larroy wrote:
> I also got a rudimentary GET command working, still is not pipelined as
> the protocol suggests.

I wouldn't worry about pipelining. It's an optimization for small
files, but in practice it will only impact the first sync.

> So I think in the following weeks/month we
> should be able to use the C++ daemon so synchronize files without
> encryption.

That will be great!

> It would be great if somebody finds the time to improve the status of
> the cryptographic parts.

I've been messing with it a little bit. The last time I looked at it I
was having a hard time deciding how to best wrap both TCP connections
and uTP connections. Perhaps it'd be better to get it to work just with
uTP to begin with.

Steven

[pe...@larroy.com]

Hi

I merged the refactor to default. Now it should be easier to implement messages from control and extensions. I separated the core functionality in src/cs/core

So to implement control.md for example, we can make a src/cs/control and copy coder.* and message.* adapted to the particular messages that need to be implemented.

I think decoupling ProtocolState and cs::core::Protocol will allow us to have a better design. Now I'm more satisfied with how messages are handled.

If you have any other ideas don't hesitate to write them on the list.

Happy hacking!

Pedro.

[Pedro Larroy]

Hi

Finished refactoring the message protocol with binary size prefixes:

https://github.com/larroy/clearskies_core/blob/master/test/protocolstate.cpp#L139


Anyone up for updating the protocol description?

Pedro

[Steven Jewel]

On 05/10/2014 02:47 PM, Pedro Larroy wrote:
>
> Finished refactoring the message protocol with binary size prefixes:
>
> https://github.com/larroy/clearskies_core/blob/master/test/protocolstate.cpp#L139
>
>
> Anyone up for updating the protocol description?

Great work! I will do it next week if nobody beats me to it.

Steven