arti: comments about catch_unwind
This commit is contained in:
parent
6b6b3d9e53
commit
5a46dfef8f
|
@ -188,6 +188,15 @@ impl TimeRounder {
|
|||
// NOTE: This function really mustn't panic. We try to log any panics
|
||||
// that we encounter, and if logging itself can panic, we're in a
|
||||
// potential heap of trouble.
|
||||
//
|
||||
// This danger is somewhat ameliorated by the behavior of the default
|
||||
// panic handler, which detects nested panics and aborts in response.
|
||||
// Thus, if we ever discard that handler, we need to be sure to
|
||||
// reimplement nested panic detection.
|
||||
//
|
||||
// Alternatively, we _could_ nest this functionality within
|
||||
// `catch_unwind`. But I'm not sure that the overhead there would be
|
||||
// acceptable: Logging can be performance sensitive.
|
||||
|
||||
use TimeRounder::*;
|
||||
/// Round `inp` down to the nearest multiple of `granularity`.
|
||||
|
|
Loading…
Reference in New Issue