Description
Applies To
Service process (could be plugins)
Describe the bug
Ran two console windows overnight as a stress test. One was constantly connecting, requesting endpoint metadata, and then disconnecting. The other simply monitoring the same device, which was also sending Active Sensing 4x a second
Over the course of 9 hours, the midisrv thread count remained unchanged, but the memory usage went from just under 4k to 18,680k. Approximately 7000k was recovered by disconnecting the listening client, dropping down to 11,020k, which then remained static and unrecovered.
Undetermined if monitoring alone is sufficient, or if it was the metadata request at fault.
Message scheduling was disabled. Translation was not in use. Only active plugin should be the metadata listener and the scheduler with disabled scheduling.
To Reproduce
Pwsh:
The looping pwsh portion looks like this:
while ($cnt -le 100000) {$cnt++; midi endpoint "\\?\SWD#MIDISRV#MIDIU_KS_BIDI_16024944077548273316_OUTPIN.0_INPIN.2#{e7cce071-3c03-423f-88d3-f1045d02552b}" request endpoint-metadata --all; Write-Host $cnt; Start-Sleep -Seconds 2 }
Dev Preview 3
Metadata
Metadata
Assignees
Labels
Type
Projects
Status
No status
Activity