Bug 154064

Summary:

REGRESSION(r196331): It made ~180 JSC tests crash on ARMv7 Linux

Product:

WebKit

Reporter:

Csaba Osztrogonác <ossy>

Component:

New Bugs

Assignee:

Csaba Osztrogonác <ossy>

Status:

RESOLVED FIXED

Severity:

Critical

CC:

barraclough, cgarcia, clopez, commit-queue, ggaren, gyuyoung.kim, hs85.jeong, keith_miller, mark.lam, msaboff, ossy, saam

Priority:

Version:

Other

Hardware:

Unspecified

OS:

Unspecified

Bug Depends on:

Bug Blocks:

108645, 154009

Attachments:

Description	Flags
Patch	none
Patch	none

Description Csaba Osztrogonác 2016-02-10 02:30:17 PST

https://trac.webkit.org/changeset/196331 made ~200 tests fail on ARMv7 Linux
(with ARM and Thumb2 instruction set too)

EFL ARMv7 Thumb2 instruction set: https://build.webkit.org/builders/EFL%20Linux%20ARMv7%20Thumb2%20Release/builds/17094
EFL ARMv7 ARM instruction set: https://build.webkit.org/builders/EFL%20Linux%20ARMv7%20Traditional%20Release/builds/16935
GTK (???) : https://build.webkit.org/builders/GTK%20Linux%20ARM%20Release/builds/10217

Comment 1 Csaba Osztrogonác 2016-02-10 07:38:35 PST

Created attachment 270989 [details]
Patch

Comment 2 Mark Lam 2016-02-10 09:19:26 PST

Comment on attachment 270989 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=270989&action=review

> Source/JavaScriptCore/bytecode/PolymorphicAccess.cpp:464
> +// EncodedJSValue in JSVALUE32_64 is a 64-bit integer. When being compiled in ARM EABI, it must be aligned even-numbered register (r0, r2 or [sp]).

By "aligned even-numbered", did you mean "aligned on an even-numbered"?

> Source/JavaScriptCore/bytecode/PolymorphicAccess.cpp:465
> +// To avoid assemblies from using wrong registers, let's occupy r1 or r3 with a dummy argument when necessary.

By "To avoid assemblies", did you mean "To prevent the assembler"?

Comment 3 Csaba Osztrogonác 2016-02-10 09:23:44 PST

I simply copy/pasted these lines from JITInlines.h (DFGSpeculativeJIT.h,
WASMFunctionCompiler.h), but I can fix the grammar everywhere, of course.

Comment 4 Csaba Osztrogonác 2016-02-10 09:30:37 PST

Created attachment 270994 [details]
Patch

Comment 5 Mark Lam 2016-02-10 09:33:48 PST

Comment on attachment 270994 [details]
Patch

r=me

Thanks for fixing the other comments as well.

Comment 6 Csaba Osztrogonác 2016-02-10 09:50:12 PST

Comment on attachment 270994 [details]
Patch

Clearing flags on attachment: 270994

Committed r196368: <http://trac.webkit.org/changeset/196368>

Comment 7 Csaba Osztrogonác 2016-02-10 09:50:19 PST

All reviewed patches have been landed.  Closing bug.