515659 - We don't handle the case where we run out of memory before we reach the soft limit properly

Status: VERIFIED FIXED

(Tamarin Graveyard :: Garbage Collection (mmGC), defect, P3)

Description

[Tommy Reilly]

We could set the soft limit to zero in this case, since its never gonna get hit anyways.

Comment 1

[Lars T Hansen]

I'm not sure what this bug is for - better description needed.  Still, targeting since it sounds important.

Comment 2

[Tommy Reilly]

The bug here is that if we go from kMemNormal to kMemReserve we will immediately issue a kMemNormal if memory is below the soft limit, we should only issue the kMemNormal if memory grew past the soft limit.  It should be as if the soft limit is ignored in this case and the normal rules for going back to kMemNormal apply

Comment 3

[Tommy Reilly]

Attachment: fix

New semantics: we only go from kMemNormal from kMemSoftLimit if the current status is kMemSoftLimit.   Additionally the means if we go from kMemNormal to kMemSoftLimit to kMemReserve we only go back to normal if we return memory from the peak, going below the soft limit won't do it.

Comment 4

[Tommy Reilly]

Comment on attachment 405324 [details] [diff] [review]
fix

Anti sent me email with approval, apparently bugzilla is giving him permissions problems.

Comment 5

[Tommy Reilly]

change 2746