Bug 211313 – REGRESSION: stress/array-buffer-view-watchpoint-can-be-fired-in-really-add-in-dfg.js fails on ARM (thumb2 only?)

• Status: NEW
• Component: JavaScriptCore

[Pablo Saavedra]

* Related to: https://bugs.webkit.org/show_bug.cgi?id=211304
* Regression appears after https://bugs.webkit.org/show_bug.cgi?id=211205

[Alexey Shvayka]

I wonder if it is https://bugs.webkit.org/show_bug.cgi?id=211020 to blame for regression? The change for ArrayBuffer is barely observable though and the test was flaky before r260992.

[Pablo Saavedra]

(In reply to Alexey Shvayka from comment #1)
> I wonder if it is https://bugs.webkit.org/show_bug.cgi?id=211020 to blame
> for regression? The change for ArrayBuffer is barely observable though and
> the test was flaky before r260992.

Yes, https://bugs.webkit.org/show_bug.cgi?id=211020 looks like a more porbable candidate.

[Pablo Saavedra]

*probable

[Alexey Shvayka]

Can I somehow obtain a crash log from EWS or reproduce the failure locally?

[Alexey Proskuryakov]

The test was skipped in https://trac.webkit.org/r261010. It was skipped for all ARM, which seems wrong, as we never had evidence of this affecting iOS.

Index: /trunk/JSTests/stress/array-buffer-view-watchpoint-can-be-fired-in-really-add-in-dfg.js
===================================================================
--- /trunk/JSTests/stress/array-buffer-view-watchpoint-can-be-fired-in-really-add-in-dfg.js	(revision 261009)
+++ /trunk/JSTests/stress/array-buffer-view-watchpoint-can-be-fired-in-really-add-in-dfg.js	(revision 261010)
@@ -1,2 +1,3 @@
+//@ skip if ["arm"].include?($architecture)
 //@ requireOptions("-e", "let iterations=40000") if ["mips"].include?($architecture)
 //@ runDefault("--jitPolicyScale=0")

[Pablo Saavedra]

(In reply to Alexey Proskuryakov from comment #5)
> The test was skipped in https://trac.webkit.org/r261010. It was skipped for
> all ARM, which seems wrong, as we never had evidence of this affecting iOS.
> 
> Index:
> /trunk/JSTests/stress/array-buffer-view-watchpoint-can-be-fired-in-really-
> add-in-dfg.js
> ===================================================================
> ---
> /trunk/JSTests/stress/array-buffer-view-watchpoint-can-be-fired-in-really-
> add-in-dfg.js	(revision 261009)
> +++
> /trunk/JSTests/stress/array-buffer-view-watchpoint-can-be-fired-in-really-
> add-in-dfg.js	(revision 261010)
> @@ -1,2 +1,3 @@
> +//@ skip if ["arm"].include?($architecture)
>  //@ requireOptions("-e", "let iterations=40000") if
> ["mips"].include?($architecture)
>  //@ runDefault("--jitPolicyScale=0")

I think I rushed myself when I skipped the test with not enough information to fill the issue. I just re-added the test again. Let's see the results for the next builds.

[Pablo Saavedra]

The test only is failing in JSCOnly Linux for ARMv7:

* https://build.webkit.org/builders/JSCOnly%20Linux%20ARMv7%20Thumb2%20SoftFP%20Release/builds/5768
* https://build.webkit.org/builders/JSCOnly%20Linux%20ARMv7%20Thumb2%20Release/builds/11484