Struct Failures 

pub struct Failures { /* private fields */ }

Every failure a device is still holding, plus the little a write scope needs around them.

One store per device rather than per stream, because a launch failing on one stream can claim a slice owned by another and both have to point at the same thing.

There is no device-wide failure beside the graph. A failure belongs to the memory it left unwritten, and work that shares no buffer with it is unaffected — which is the whole point, and a slot that failed the next flush regardless of what it was flushing was the one thing that broke it.

Implementations

impl Failures

pub fn new(logger: Arc<ServerLogger>) -> Self

An empty store for a device that has failed at nothing yet.

pub fn graph(&self) -> &ErrorGraph

What the failure ids carried by this device’s allocations mean.

pub fn graph_mut(&mut self) -> &mut ErrorGraph

graph mutably, handed down to every reserve, bind and cleanup — those are where slices shed the failures they carry.

Trait Implementations

impl Debug for Failures

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.