[QM_TRY] Track nested error propagation
Categories
(Core :: Storage: Quota Manager, enhancement)
Tracking
()
People
(Reporter: jstutte, Unassigned)
References
(Depends on 1 open bug, Blocks 2 open bugs)
Details
What we actually want to determine through our heuristics in bug 1700915 is the nesting level of QM_TRY and friends. Since we are looking at “propagation stacks”, it is natural to assume that this level can be bound to the error value itself. We could just count the number of QM_* macros that passed along this error.
This could be achieved by a common wrapper class around the errors. And it would be kind of the inverse variant of something similar we did for SpinEventLoopUntil, but being more loosely coupled (looking at error propagation as such, not real, compiler scoped function call nesting).
|
||
Comment 1•3 years ago
|
||
Ah, sorry about filing a new bug.
|
Reporter | |
Updated•3 years ago
|
|
|
Reporter | |
Comment 2•3 years ago
|
||
We might want to expand this a bit (like buffering the propagation stack in memory before sending it via telemetry in case of sure error).
|
|
||
Comment 3•3 years ago
|
||
(In reply to Jens Stutte [:jstutte] from comment #2)
We might want to expand this a bit (like buffering the propagation stack in memory before sending it via telemetry in case of sure error).
Yeah, that might be possible, but it will require bigger changes.
|
|
||
Comment 4•3 years ago
|
||
(In reply to Jan Varga [:janv] from comment #3)
(In reply to Jens Stutte [:jstutte] from comment #2)
We might want to expand this a bit (like buffering the propagation stack in memory before sending it via telemetry in case of sure error).
Yeah, that might be possible, but it will require bigger changes.
Maybe not, but I guess buffering would only improve performance which seems to be ok at the moment.
|
|
Reporter | |
Comment 5•3 years ago
|
||
For now I think bug 1709352 is more than enough. But we might want to move towards client side aggregation of equal propagation stacks in order to reduce telemetry significantly (and maybe have it even in release). And a first step for this would be to have a representation of a propagation stack in-memory.
Description
•