Bug 142629

Summary:	REGRESSION(r181326): It made dfg-int16array.js and dfg-int8array.js fail on AArch64 Linux
Product:	WebKit	Reporter:	Csaba Osztrogonác <ossy>
Component:	JavaScriptCore	Assignee:	Nobody <webkit-unassigned>
Status:	RESOLVED DUPLICATE
Severity:	Normal	CC:	andersca, cgarcia, fpizlo, mcatanzaro, msaboff, ossy
Priority:	P2
Version:	528+ (Nightly build)
Hardware:	Unspecified
OS:	Unspecified
See Also:	https://bugs.webkit.org/show_bug.cgi?id=138390
Bug Depends on:
Bug Blocks:	108645, 142524

Description Csaba Osztrogonác 2015-03-12 10:49:58 PDT

http://trac.webkit.org/changeset/181326 made the following tests fail on AArch64 Linux:
FAIL: jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit
FAIL: jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-no-llint
FAIL: jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-ftl-eager-no-cjit
FAIL: jsc-layout-tests.yaml/js/script-tests/dfg-int8array.js.layout-dfg-eager-no-cjit
FAIL: jsc-layout-tests.yaml/js/script-tests/dfg-int8array.js.layout-no-llint
FAIL: jsc-layout-tests.yaml/js/script-tests/dfg-int8array.js.layout-ftl-eager-no-cjit

jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: DIFF FAILURE!
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: --- ../.tests/jsc-layout-tests.yaml/js/dfg-int16array-expected.txt    2015-03-12 09:21:28.000000000 -0700
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: +++ ../jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit.out   2015-03-12 10:36:15.100000000 -0700
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: @@ -5126,12 +5126,12 @@
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit:  PASS getter(a, 20, 30840) is 30840
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit:  PASS getter(a, 21, -31097) is -31097
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit:  PASS getter(a, 22, 31354) is 31354
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: -PASS getter(a, 23, -31611) is -31611
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: +FAIL getter(a, 23, -31611) should be -31611. Was -31611.
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit:  PASS getter(a, 24, 31868) is 31868
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: -PASS getter(a, 25, -32125) is -32125
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: +FAIL getter(a, 25, -32125) should be -32125. Was -32125.
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit:  PASS getter(a, 26, 32382) is 32382
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: -PASS getter(a, 27, -32639) is -32639
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: -PASS getter(a, 28, 32896) is -32640
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: +FAIL getter(a, 27, -32639) should be -32639. Was -32639.
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit: +FAIL getter(a, 28, 32896) should be -32640. Was -32640.
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit:  PASS getter(a, 29, -33153) is 32383
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit:  PASS getter(a, 30, 33410) is -32126
jsc-layout-tests.yaml/js/script-tests/dfg-int16array.js.layout-dfg-eager-no-cjit:  PASS getter(a, 31, -33667) is 31869

All of these failure is similar, X != X, how is it possible?

Comment 1 Carlos Garcia Campos 2015-03-12 11:02:38 PDT

+FAIL getter(a, 23, -31611) should be -31611. Was -31611.

Weird.

Comment 2 Michael Catanzaro 2015-03-13 13:22:29 PDT

I have no clue. The patch contained no functional change, but if a function marked NORETURN has a return statement then things could get wonky. But there would definitely be a compiler warning about that.

Comment 3 Csaba Osztrogonác 2015-03-16 06:35:03 PDT

It seems the patch in bug142575 fixes this issue. (or fixed by other patch in trunk nowadays - will check soon)

Comment 4 Csaba Osztrogonác 2015-03-16 06:37:09 PDT

(In reply to comment #3)
> It seems the patch in bug142575 fixes this issue. (or fixed by other patch
> in trunk nowadays - will check soon)

Argh, I forgot to revert the local revert of r181326. :)
Please ignore this comment, the bug is still valid on trunk.

Comment 5 Csaba Osztrogonác 2015-03-26 03:05:58 PDT

I'm going to skip it on Linux too, these tests fail on iOS too, but 
we don't know if it is the same bug or not. See bug138390 for details.

Comment 6 Michael Saboff 2015-03-27 16:57:27 PDT

(In reply to comment #5)
> I'm going to skip it on Linux too, these tests fail on iOS too, but 
> we don't know if it is the same bug or not. See bug138390 for details.

I really think this is a dup of bug138390.  Preparing a patch for that bug now.

Comment 7 Michael Saboff 2015-03-27 17:43:53 PDT

Landed fix for bug138390 in change set r182091: <http://trac.webkit.org/changeset/182091>.  I kept the tests as skipped for linux for someone with a linux system to verify that my fix takes care of this issue as well.

Comment 8 Csaba Osztrogonác 2015-03-28 03:00:08 PDT

(In reply to comment #7)
> Landed fix for bug138390 in change set r182091:
> <http://trac.webkit.org/changeset/182091>.  I kept the tests as skipped for
> linux for someone with a linux system to verify that my fix takes care of
> this issue as well.

Thanks for the fix, I checked it on Linux, these test pass now, so 
I already unskipped them - http://trac.webkit.org/changeset/182110 .

*** This bug has been marked as a duplicate of bug 138390 ***