Bug 201986 - JSTests/stress/ftl-put-by-id-setter-exception-interesting-live-state.js ftl-eager flavor fails intermittently.
Summary: JSTests/stress/ftl-put-by-id-setter-exception-interesting-live-state.js ftl-e...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Tadeu Zagallo
URL:
Keywords: InRadar
: 183266 (view as bug list)
Depends on:
Blocks:
 
Reported: 2019-09-19 09:20 PDT by Mark Lam
Modified: 2019-09-22 11:39 PDT (History)
8 users (show)

See Also:


Attachments
Patch (19.70 KB, patch)
2019-09-20 17:29 PDT, Tadeu Zagallo
no flags Details | Formatted Diff | Diff
Patch (20.61 KB, patch)
2019-09-20 18:55 PDT, Tadeu Zagallo
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Mark Lam 2019-09-19 09:20:12 PDT
This is how I reproduce this failure with a release build on trunk r250085:

$ while true; do VM=WebKitBuild/Release && DYLD_FRAMEWORK_PATH=$VM $VM/jsc --airForceBriggsAllocator=true --forcePolyProto=true --thresholdForJITAfterWarmUp=10 --thresholdForJITSoon=10 --thresholdForOptimizeAfterWarmUp=20 --thresholdForOptimizeAfterLongWarmUp=20 --thresholdForOptimizeSoon=20 --thresholdForFTLOptimizeAfterWarmUp=20 --thresholdForFTLOptimizeSoon=20 --thresholdForOMGOptimizeAfterWarmUp=20 --thresholdForOMGOptimizeSoon=20 --maximumEvalCacheableSourceLength=150000 --useEagerCodeBlockJettisonTiming=true --collectContinuously=true --useGenerationalGC=false JSTests/stress/ftl-put-by-id-setter-exception-interesting-live-state.js; done

When the test fails, it prints the following to stdout:
Exception: Error42
Comment 1 Radar WebKit Bug Importer 2019-09-19 09:20:35 PDT
<rdar://problem/55521953>
Comment 2 Mark Lam 2019-09-19 09:23:49 PDT
I forgot to say that in order to make this issue reproduce more quickly, I kicked off a full WebKit build in the background (separate terminal window) to introduce some noise into the system.
Comment 3 Tadeu Zagallo 2019-09-20 17:29:32 PDT
Created attachment 379293 [details]
Patch
Comment 4 Saam Barati 2019-09-20 17:43:23 PDT
Comment on attachment 379293 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=379293&action=review

r=me

> Source/JavaScriptCore/bytecode/AccessCase.cpp:238
> +bool AccessCase::doesCalls(Vector<JSCell*>* cellsToMarkIfDoesCalls) const

let's implement Yusuke's idea of listing all types() as a switch both in here and in forEachDependentCell, so in the future, when someone adds a new type, they'll get a compile error unless they implement the type in these functions.

> Source/JavaScriptCore/bytecode/PropertyCondition.cpp:354
>  bool PropertyCondition::isStillLive(VM& vm) const

If this is no longer called, let's remove it

> Source/JavaScriptCore/jit/PolymorphicCallStubRoutine.cpp:132
> +    bool isValid = true;

nit: let's call this isStillLive
Comment 5 Yusuke Suzuki 2019-09-20 17:44:50 PDT
r=me too, nice!
Comment 6 Tadeu Zagallo 2019-09-20 18:55:13 PDT
Created attachment 379301 [details]
Patch
Comment 7 WebKit Commit Bot 2019-09-21 11:30:35 PDT
Comment on attachment 379301 [details]
Patch

Clearing flags on attachment: 379301

Committed r250184: <https://trac.webkit.org/changeset/250184>
Comment 8 WebKit Commit Bot 2019-09-21 11:30:37 PDT
All reviewed patches have been landed.  Closing bug.
Comment 9 Tadeu Zagallo 2019-09-22 11:39:33 PDT
*** Bug 183266 has been marked as a duplicate of this bug. ***