Open Bug 1509032 Opened 3 years ago Updated 3 years ago
The pingsender leaves behind zombie processes on Linux
While using nightly with telemetry enabled I noticed that multiple pingsender instances had zombie processes left behind. This is peculiar because it points to a flaw in the nsIProcess implementation on Linux. When launching pingsender we use nsIProcess.run() in non-blocking mode which spanws a thread to wait for the generated process: https://searchfox.org/mozilla-central/rev/55895c49f55073d82d977cb74ec1d3a71ae4b25f/xpcom/threads/nsProcessCommon.cpp#576 This will eventually invoke nsProcess::Monitor() which in turn waitpid()s for the process: https://searchfox.org/mozilla-central/rev/55895c49f55073d82d977cb74ec1d3a71ae4b25f/xpcom/threads/nsProcessCommon.cpp#272 But apparently something goes amiss. For now I'm filing this in telemetry but I'll move it in XPCOM if it's really a defect in nsIProcess instead of on how we launch pingsender.
Those links seem to be cut off.
It's a BMO issue apparently (saw people talk about it on #developers). The first link is to xpcom/threads/nsProcessCommon.cpp:576 and the second to xpcom/threads/nsProcessCommon.cpp:272
Are you looking into this issue currently Gabriele? Or is this filed for future investigation?
I will be looking into this ASAP.
Ok, let us know if you need anything from us.
Assignee: nobody → gsvelto
Priority: -- → P2
You need to log in before you can comment on or make changes to this bug.