Closed
Bug 1209779
Opened 9 years ago
Closed 9 years ago
Bug 1204584 broke the Gecko Profiler
Categories
(Core :: JavaScript Engine, defect)
Tracking
()
RESOLVED
FIXED
mozilla44
Tracking | Status | |
---|---|---|
firefox42 | --- | unaffected |
firefox43 | + | fixed |
firefox44 | --- | fixed |
People
(Reporter: vladan, Assigned: fitzgen)
References
Details
(Keywords: regression)
Attachments
(1 file)
1.08 KB,
patch
|
shu
:
review+
lizzard
:
approval-mozilla-aurora+
|
Details | Diff | Splinter Review |
Bug 1204584 broke the Gecko Profiler (it never returns from retrieving the profile). Nightly 44 and Aurora 43 are affected I used mozRegression to find the offending changeset: https://hg.mozilla.org/integration/mozilla-inbound/rev/f33142322571 STR: 1. Set xpinstall.signatures.required to false (needed for next step) 2. Install the unsigned Gecko Profiler extension from https://github.com/bgirard/Gecko-Profiler-Addon/raw/master/src/geckoprofiler.xpi 3. Click Analyze to show a profile Expected: Profile opens Actual: Profiler never returns from "Retrieving profile"
Reporter | ||
Updated•9 years ago
|
Comment 1•9 years ago
|
||
I get the following error when using Gecko Profiler: TypeError: types is undefined src/main.js:866:9 Looks like when inflating the optimizations in the frame, the value in the optimizations slot is a number, not an object containing sites and attempts, etc., not sure how that patch could've caused that, or if it's from something else and not evident in the devtools performance tool (even in browser toolbox)
Updated•9 years ago
|
Flags: needinfo?(nfitzgerald)
Assignee | ||
Comment 2•9 years ago
|
||
I'll look into this.
Assignee: nobody → nfitzgerald
Status: NEW → ASSIGNED
Flags: needinfo?(nfitzgerald)
Assignee | ||
Comment 3•9 years ago
|
||
Was able to reproduce the type error:
> TypeError: types is undefined main.js:866:9
Assignee | ||
Comment 4•9 years ago
|
||
If I revert the patch from bug 1204854 locally, I start hitting assertion failures: Assertion failure: entry->isMarkedFromAnyThread(rt), at /Users/fitzgen/src/mozilla-central/js/src/jit/JitcodeMap.cpp:484 [Child 73545] ###!!! ABORT: Aborting on channel error.: file /Users/fitzgen/src/mozilla-central/ipc/glue/MessageChannel.cpp, line 1768 #01: mozilla::ipc::MessageChannel::OnChannelErrorFromLink()[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x892ca4] #02: mozilla::ipc::ProcessLink::OnChannelError()[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x8946a1] #03: non-virtual thunk to mozilla::ipc::ProcessLink::OnChannelError()[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x8946cc] #04: IPC::Channel::ChannelImpl::OnFileCanReadWithoutBlocking(int)[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x812e3c] #05: base::MessagePumpLibevent::OnLibeventNotification(int, short, void*)[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x7c51fe] #06: event_persist_closure[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x873efa] #07: event_process_active_single_queue[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x873750] #08: event_process_active[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x857fdc] #09: event_base_loop[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x8570d7] #10: base::MessagePumpLibevent::Run(base::MessagePump::Delegate*)[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x7c563c] #11: MessageLoop::RunInternal()[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x7c3125] #12: MessageLoop::RunHandler()[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x7c3035] #13: MessageLoop::Run()[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x7c2fdd] #14: base::Thread::ThreadMain()[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x7e8436] #15: ThreadFunc(void*)[/Users/fitzgen/src/mozilla-central/obj.noindex/dist/NightlyDebug.app/Contents/MacOS/XUL +0x7e960e] #16: _pthread_body[/usr/lib/system/libsystem_pthread.dylib +0x3268] #17: _pthread_body[/usr/lib/system/libsystem_pthread.dylib +0x31e5] [Child 73545] ###!!! ABORT: Aborting on channel error.: file /Users/fitzgen/src/mozilla-central/ipc/glue/MessageChannel.cpp, line 1768 Hit MOZ_CRASH() at /Users/fitzgen/src/mozilla-central/memory/mozalloc/mozalloc_abort.cpp:33
Comment 5•9 years ago
|
||
I reverted the patch locally as well to use the profiler and haven't seen this assertion/crash.
Comment 6•9 years ago
|
||
(In reply to Kartikaya Gupta (email:kats@mozilla.com) from comment #5) > I reverted the patch locally as well to use the profiler and haven't seen > this assertion/crash. It's a sanity assertion, probably not a crash. Are you running a debug build? If not, you probably won't see this.
Comment 7•9 years ago
|
||
Yup, running a debug build. I didn't see the assertion stack either but it's possibly it flew by in the console output while I wasn't looking. I was getting a lot of other spew on the console so I might have missed it. It's certainly not hitting frequently though.
Assignee | ||
Comment 8•9 years ago
|
||
Interesting, in debug builds it should crash the browser.
Assignee | ||
Comment 9•9 years ago
|
||
Attachment #8668159 -
Flags: review?(shu)
Assignee | ||
Comment 10•9 years ago
|
||
Turned out to be a latent bug exposed by my patch. Try push: https://treeherder.mozilla.org/#/jobs?repo=try&revision=83333509c891
Assignee | ||
Comment 11•9 years ago
|
||
Not sure if this goes without saying, but I verified that the GP addon works with this patch applied.
Comment 12•9 years ago
|
||
Comment on attachment 8668159 [details] [diff] [review] Ensure that all null elements are written when streaming profiler JSON Review of attachment 8668159 [details] [diff] [review]: ----------------------------------------------------------------- Sorry about this, thanks.
Attachment #8668159 -
Flags: review?(shu) → review+
Assignee | ||
Updated•9 years ago
|
Keywords: checkin-needed
Comment 13•9 years ago
|
||
https://hg.mozilla.org/integration/mozilla-inbound/rev/ce8c63fd5fec
Keywords: checkin-needed
Comment 16•9 years ago
|
||
https://hg.mozilla.org/mozilla-central/rev/ce8c63fd5fec
Status: ASSIGNED → RESOLVED
Closed: 9 years ago
status-firefox44:
--- → fixed
Resolution: --- → FIXED
Target Milestone: --- → mozilla44
Reporter | ||
Comment 17•9 years ago
|
||
Comment on attachment 8668159 [details] [diff] [review] Ensure that all null elements are written when streaming profiler JSON Approval Request Comment [Feature/regressing bug #]: 1204584 [User impact if declined]: Impossible to profile performance problems on Firefox 43 [Describe test coverage new/current, TreeHerder]: [Risks and why]: None, tiny fix, confined to profiler code [String/UUID change made/needed]: None
Attachment #8668159 -
Flags: approval-mozilla-aurora?
Comment 18•9 years ago
|
||
The patch that landed was empty. ni? cbook so he knows whatever process he used to land it (comment 13) doesn't work. I relanded it on inbound.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Comment 22•9 years ago
|
||
https://hg.mozilla.org/mozilla-central/rev/39983a1913cf
Status: REOPENED → RESOLVED
Closed: 9 years ago → 9 years ago
Resolution: --- → FIXED
Comment 23•9 years ago
|
||
Comment on attachment 8668159 [details] [diff] [review] Ensure that all null elements are written when streaming profiler JSON Minor fix for profiler regression; let's uplift to aurora.
Attachment #8668159 -
Flags: approval-mozilla-aurora? → approval-mozilla-aurora+
Updated•9 years ago
|
status-firefox43:
--- → affected
tracking-firefox43:
--- → +
Updated•9 years ago
|
status-firefox42:
--- → unaffected
You need to log in
before you can comment on or make changes to this bug.
Description
•