Skip to content

Latest commit

 

History

History
115 lines (83 loc) · 4.25 KB

FAQ.md

File metadata and controls

115 lines (83 loc) · 4.25 KB
Raw

Frequently Asked Questions

How To Send Data Where The Size Is Unknown At Compilation-Time?

Take a look at the publish-subscribe dynamic data size example.

The idea is to create a service based on a slice and define at runtime a max_slice_len. Then samples up to a length of the max slice length can be allocated with loan_slice{_uninit}. When it turns out that the slice length is insufficient, a new publisher with a larger max_slice_len can be created.

Important

Be aware that the history of the old publisher is lost when it is removed.

Note

We are also working on an API that does not require the user to explicitly create a new publisher whenever the memory is insufficient. It would also solve the history issue.

How To Make 32-bit and 64-bit iceoryx2 Applications Interoperatable

Use the feature flag enforce_32bit_rwlock_atomic which enforces 32-bit atomics for all targets at the cost of the lock-free guarantee. Meaning, when an application crashes at the wrong point in time it can lead to a system deadlock.

My Transmission Type Is Too Large, Encounter Stack Overflow On Initialization.

Take a look at the complex data types example.

In this example the PlacementDefault trait is introduced that allows in place initialization and solves the stack overflow issue when the data type is larger than the available stack size.

Does iceoryx2 Offer an Async API?

No, but it is on our roadmap. However, we offer an event-based API to implement push notifications. For more details, see the event example.

Application does not remove services/ports on shutdown or several application restarts lead to port count exceeded

The structs of iceoryx2 need to be able to cleanup all resources when they go out of scope. This is not the case when the application is:

  • killed with the sigkill signal (kill -9)
  • the SIGTERM signal is not explicitly handled

iceoryx2 already provides a mechanism that registers a signal handler that handles termination requests gracefully, see publish subscribe example and

while let Iox2Event::Tick = Iox2::wait(CYCLE_TIME) {
  // user code
}

But you can also use a crate like ctrlc.

How to use log or tracing as default log backend

  • log, add the feature flag logger_log to the dependency in Cargo.toml
iceoryx2 = { version = "0.1.0", features = ["logger_log"]}
  • tracing, add the feature flag logger_tracing to the dependency in Cargo.toml
 iceoryx2 = { version = "0.1.0", features = ["logger_tracing"]}

A crash leads to the failure PublishSubscribeOpenError(UnableToOpenDynamicServiceInformation)

Note: A command line tool and internal service is already planned to cleanup resources from crashed applications, see issue #65.

When an application crashes some resources may remain in the system and must be cleaned up manually. If this occurs, stop all services and remove manually all shared memory segments and static service config files.

rm -rf /dev/shm/iox2_*
rm -rf /tmp/iceoryx2/*

If you cannot stop all running services, you can look up the uuid of the service in question and remove the files manually. Assume, the service My/Funk/ServiceName is corrupted. You can identify the static config by grepping the service name in the /tmp/iceoryx2/service folder.

So the command

cd /tmp/iceoryx2/service
grep -RIne "My/Funk/ServiceName"

provides us with the output

iox2_25b25afeb7557886e9f69408151e018e268e5917.service:2:service_name = "My/Funk/ServiceName"

The file name corresponds with the uuid of the service. So removing the dynamic and static service config with the following commands, removes the service completely from the system.

# static service config
rm /tmp/iceoryx2/service/iox2_25b25afeb7557886e9f69408151e018e268e5917.service

# dynamic service config
rm /dev/shm/iox2_25b25afeb7557886e9f69408151e018e268e5917.dynamic

Be aware, that if an application with a publisher crashes, the data segment of the publisher must be cleaned up as well.