Bugzilla

So after a bit of experimentation I could use a bit of assistance on how to go about this.

When I try to accumulate a value too big for an unsigned 64 bit int (2^65), the buckets that show up are not what we _should_ see if this were defined behavior. Instead the tail of the distro reads: ["8760003",0],["9147841",0],["9552851",0],["9975792",1],["10417458",0]]

If I understand, the final bucket should have the logged value or the key should be over 2^65, correct? In fact, it will do this for much smaller values (e.g. 2^61). Ok well it is out of range, but I'd have expected an exception.

As it happens, 2^35 throws.

I did notice that [values above 1tb/2^12 should cause an invaliderror to be logged](https://mozilla.github.io/glean/book/reference/metrics/memory_distribution.html#recorded-errors). But, there is no Desktop testGetNumRecordedErrors.

I'm not sure how to proceed here. If you like, I can file a bug for the test interface and start implementing that so we can count, I could write a test that generically confirms this doesn't cause the value to go into the final bucket, we could say that the value is clamped and just let it go and I could close, or I'm open to suggestions if you don't like any of these.

So after a bit of experimentation I could use a bit of assistance on how to go about this.

When I try to accumulate a value too big for an unsigned 64 bit int (2^65), the buckets that show up are not what (I would expect) we _should_ see if this were defined behavior. Instead the tail of the distro reads: ["8760003",0],["9147841",0],["9552851",0],["9975792",1],["10417458",0]]

If I understand, the final bucket should have the logged value or the key should be over 2^65, correct? In fact, it will do this for much smaller values (e.g. 2^61). Ok well it is out of range, but I'd have expected an exception.

As it happens, 2^35 throws.

I did notice that [values above 1tb/2^12 should cause an invaliderror to be logged](https://mozilla.github.io/glean/book/reference/metrics/memory_distribution.html#recorded-errors). But, there is no Desktop testGetNumRecordedErrors.

I'm not sure how to proceed here. If you like, I can file a bug for the test interface and start implementing that so we can count, I could write a test that generically confirms this doesn't cause the value to go into the final bucket, we could say that the value is clamped and just let it go and I could close, or I'm open to suggestions if you don't like any of these.