Adds support for asio::cancel_after #324
Conversation
|
This may still yield CI errors due to the race conditions fixed by #320, so this one should be merged before this. |
|
|
||
| In this situation, we don't know if the request was processed by the server or not. | ||
| And we have no way to know it. By default, the library mark these requests as | ||
| failed with `asio::error::operation_aborted`. (TODO: do we want another error code here?). |
There was a problem hiding this comment.
the library mark these
Needs an s at the end.
do we want another error code here?
I think we do, otherwise the user might not know why it was canceled. Transforming operation_aborted into e.g. unresponded_on_connection_lost will make it clear.
| req.get_config().cancel_if_unresponded = false; // retry | ||
|
|
||
| // Makes sure that the key is set, even in the presence of network errors. | ||
| // This may suspend for an unspecified period of time if the server is down. |
There was a problem hiding this comment.
This may suspend for an unspecified period of time if the server is down.
I would rephrase this to something like
This will remain suspended until the server is capable of serving requests again e.g. after a process manager restarted the server.
|
|
||
| private: | ||
| // Function object to initiate the async ops that use asio::any_completion_handler. | ||
| // Required for asio::cancel_after to work. |
There was a problem hiding this comment.
Why is this required? i.e. why wouldn't the lambda be enough?
There was a problem hiding this comment.
cancel_after (without a timer argument) requires an executor to construct a temporary timer for the timeout. This executor is taken from the initiation object. This gets done for free if you use async_compose, but needs to be done manually when using async_initiate.
cancel_if_unrespondedflagclose #226