- This patch was written in 3 interdependent parts, which are described below -
Part 1A: Allow HangStack to contain raw PCs and Module offsets, r=froydnj
The HangStack previously consisted of an array of const char* pointers into its
backing string buffer, which represented pseudostack entries. With interleaved
stacks, it is now possible for the stack to contain raw unresolved program
counters (Kind::PC), and module/offset pairs (Kind::MODOFFSET). To do this, we
use a discriminated union, and make the backing array use the discriminated
union instead of const char*s.
The code cannot use mozilla::Variant<const char*, uintptr_t, Module>
unfortuantely, as we cannot use the implementation of ParamTraits for Variant in
HangStack's ParamTraits implementation.
When deserializing a HangStack over IPC, we need to read the string frame
entries into the backing string buffer, and generate const char* entries for
each of the strings which we read in over IPC. The default implementation of
ParamTraits wouldn't give us access to the enclusing HangStack object while
deserializing each individual entry, so we couldn't use it. In fact, Entries
don't have ParamTraits implemented for them at all, and can only be sent over
IPC as part of a HangStack due to this dependency.
Part 1B: Remove nsIHangDetails.pseudoStack, replace ProcessedStack w/ new HangStack type, r=froydnj
Previously there were two stack objects on each HangDetails object: mStack and
mPseudoStack. mStack was a Telemetry::ProcessedStack, while mPseudoStack was a
HangStack. After the changes in part 1A, HangStack can now contain all of the
information of both the old HangStack and ProcessedStack, so the mPseudoStack
field is renamed to mStack, and the old mStack field is removed.
This patch also implements the new GetStack getter, which generates the JS data
format for the new HangStack type.
Part 1C: Collect interleaved stacks w/ ProfilerStackCollector API in ThreadStackHelper, r=froydnj
This new API was added by njn in bug 1380286, and provides both pseudostack and
native stack entries to the consumer of the API.
This patch changes ThreadStackHelper to use this new API instead of the previous
one, and use it to collect the frames directly into HangStack objects.
HangAnnotations was very complex, required a separate allocation, and used this
unfortunate virtual interface implementation which made it harder to do
interesting things with it (such as serialize it over IPC).
This new implementation is much simpler and more concrete, making
HangAnnotations simply be a nsTArray<Annotation>. This also simplifies some of
the IPC code which was added in part 7.
MozReview-Commit-ID: EzaaxdHpW1t
These changes are going to increase the amount of data which we collect from BHR
a lot. It would be dangerous to run it on beta, especially considering how soon
the next merge is.
This should turn it off for 100% of beta users if I understand the logic
correctly.
MozReview-Commit-ID: 3HyEKWdXaqU
BHRTelemetryService only runs in the parent process (and we can only submit
pings from there), so we need to send the data which we collect in the GPU and
Content processes over IPC to the parent process.
MozReview-Commit-ID: 8B5uZKbjNbU
This patch adds the BHRTelemetryService which is a JS implemented XPCOM service
that simply listens to the bhr-thread-hang observer notification, and uses the
data it collects from it to submit telemetry pings.
MozReview-Commit-ID: 2hPXAFmHrm5
We're going to use HangDetails as the type containing hang information. We'll
have a JS component which reads the data out of nsIHangDetails, builds the
payload, and submits it to telemetry for us.
We'll do it in JS because telemetry has to be submitted from JS.
This patch also adds IPC serization for the relevant types so that we can send
HangDetails objects over IPDL.
MozReview-Commit-ID: CeikKabY9Vs
HangAnnotations was very complex, required a separate allocation, and used this
unfortunate virtual interface implementation which made it harder to do
interesting things with it (such as serialize it over IPC).
This new implementation is much simpler and more concrete, making
HangAnnotations simply be a nsTArray<Annotation>. This also simplifies some of
the IPC code which was added in part 7.
MozReview-Commit-ID: EzaaxdHpW1t
These changes are going to increase the amount of data which we collect from BHR
a lot. It would be dangerous to run it on beta, especially considering how soon
the next merge is.
This should turn it off for 100% of beta users if I understand the logic
correctly.
MozReview-Commit-ID: 3HyEKWdXaqU
BHRTelemetryService only runs in the parent process (and we can only submit
pings from there), so we need to send the data which we collect in the GPU and
Content processes over IPC to the parent process.
MozReview-Commit-ID: 8B5uZKbjNbU
This patch adds the BHRTelemetryService which is a JS implemented XPCOM service
that simply listens to the bhr-thread-hang observer notification, and uses the
data it collects from it to submit telemetry pings.
MozReview-Commit-ID: 2hPXAFmHrm5
We're going to use HangDetails as the type containing hang information. We'll
have a JS component which reads the data out of nsIHangDetails, builds the
payload, and submits it to telemetry for us.
We'll do it in JS because telemetry has to be submitted from JS.
This patch also adds IPC serization for the relevant types so that we can send
HangDetails objects over IPDL.
MozReview-Commit-ID: CeikKabY9Vs
Michael Layzell
Sebastian Hengst