1216151 - Assertion failure: !hasLazyGroup(), at js/src/jsobj.h:135

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Christian Holler (:decoder)]

The following testcase crashes on mozilla-central revision d1a89632277f (build with --enable-optimize --enable-posix-nspr-emulation --enable-valgrind --enable-gczeal --disable-tests --enable-debug, run with --ion-eager --ion-offthread-compile=off):

setJitCompilerOption("baseline.warmup.trigger", 10)
loadFile(`
var o = {}
for (var i = 0; i < 10; ++i)
  o.x = ""
`);
function loadFile(lfVarx) eval("(function() { " + lfVarx + " })()")



Backtrace:

Program received signal SIGSEGV, Segmentation fault.
0x000000000044286c in JSObject::group (this=0x7ffff7e620a0) at js/src/jsobj.h:135
#0  0x000000000044286c in JSObject::group (this=0x7ffff7e620a0) at js/src/jsobj.h:135
#1  0x000000000064c0e4 in group (this=this@entry=0x7ffff7e620a0) at js/src/vm/ObjectGroup.h:118
#2  CanInlineSetPropTypeCheck (obj=obj@entry=0x7ffff7e620a0, id=..., val=..., checkTypeset=checkTypeset@entry=0x7fffffffa6d0) at js/src/jit/IonCaches.cpp:3064
#3  0x000000000067d336 in IsPropertySetInlineable (checkTypeset=0x7fffffffa6d0, needsTypeBarrier=<optimized out>, val=..., pshape=..., id=..., obj=0x7ffff7e620a0) at js/src/jit/IonCaches.cpp:3119
#4  CanAttachNativeSetProp (checkTypeset=0x7fffffffa6d0, shape=..., holder=..., needsTypeBarrier=<optimized out>, val=..., id=..., obj=..., cx=0x7ffff6907400) at js/src/jit/IonCaches.cpp:3200
#5  js::jit::SetPropertyIC::tryAttachNative (this=this@entry=0x7ffff69a94f0, cx=cx@entry=0x7ffff6907400, outerScript=..., outerScript@entry=..., ion=ion@entry=0x7ffff69a9400, obj=..., obj@entry=..., id=..., id@entry=..., emitted=emitted@entry=0x7fffffffa8e0, tryNativeAddSlot=tryNativeAddSlot@entry=0x7fffffffa8f0) at js/src/jit/IonCaches.cpp:3412
#6  0x000000000067f060 in js::jit::SetPropertyIC::tryAttachStub (this=this@entry=0x7ffff69a94f0, cx=cx@entry=0x7ffff6907400, outerScript=outerScript@entry=..., ion=ion@entry=0x7ffff69a9400, obj=obj@entry=..., id=..., id@entry=..., emitted=emitted@entry=0x7fffffffa8e0, tryNativeAddSlot=tryNativeAddSlot@entry=0x7fffffffa8f0) at js/src/jit/IonCaches.cpp:3474
#7  0x000000000067f28a in js::jit::SetPropertyIC::update (cx=0x7ffff6907400, outerScript=..., cacheIndex=<optimized out>, obj=..., value=...) at js/src/jit/IonCaches.cpp:3539
#8  0x00007ffff7fee03d in ?? ()
#9  0xff000000ff000000 in ?? ()
#10 0x00007fffffffaa68 in ?? ()
#11 0x0000000000000000 in ?? ()
rax	0x0	0
rbx	0x7ffff7e5d340	140737352422208
rcx	0x7ffff6ca53cd	140737333842893
rdx	0x0	0
rsi	0x7ffff6f7a9d0	140737336814032
rdi	0x7ffff6f791c0	140737336807872
rbp	0x7fffffffa600	140737488332288
rsp	0x7fffffffa600	140737488332288
r8	0x7ffff7fe0780	140737354008448
r9	0x6372732f736a2f6c	7165916604736876396
r10	0x7fffffffa3c0	140737488331712
r11	0x7ffff6c27960	140737333328224
r12	0x7ffff69a94f0	140737330713840
r13	0x7fffffffa700	140737488332544
r14	0x7ffff7e620a0	140737352442016
r15	0x7fffffffa6e0	140737488332512
rip	0x44286c <JSObject::group() const+28>
=> 0x44286c <JSObject::group() const+28>:	movl   $0x87,0x0
   0x442877 <JSObject::group() const+39>:	callq  0x4a56f0 <abort()>

Comment 1

[Fuzzing Team]

JSBugMon: Bisection requested, result:
=== Treeherder Build Bisection Results by autoBisect ===

The "good" changeset has the timestamp "20151015065148" and the hash "e261ab579dfb744fa1a9cd72eb36054d364907c4".
The "bad" changeset has the timestamp "20151015065349" and the hash "0ea76f0263ef67182b05fb12970ad24136365584".

Likely regression window: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=e261ab579dfb744fa1a9cd72eb36054d364907c4&tochange=0ea76f0263ef67182b05fb12970ad24136365584

Comment 2

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

new function() {
    var x = {};
    p(function() {
        x;
    })
}

is another testcase that asserts identically with --fuzzing-safe --no-threads --ion-eager on m-c rev e8c7dfe727cd.

Comment 3

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

autoBisect shows this is probably related to the following changeset:

The first bad revision is:
changeset:   https://hg.mozilla.org/mozilla-central/rev/03f521f50792
user:        Jan de Mooij
date:        Thu Oct 15 15:52:42 2015 +0200
summary:     Bug 1214163 - Clean up SetPropertyIC::update. r=efaust

Jan, is bug 1214163 a likely regressor?

Comment 4

[Jan de Mooij [:jandem]]

Attachment: Patch

Before the refactoring, SetPropertyIC::update happened to call obj->getGroup before trying to attach a stub. Some code there depended on this by expecting a non-lazy group.

The simplest fix is to move the getGroup call before the tryAttachStub call.

We could also pass a JSContext down to CanInlineSetPropTypeCheck or return false there if we have a lazy group, those options didn't seem great either.

Comment 5

[Eric Faust [:efaust]]

Comment on attachment 8675799 [details] [diff] [review]
Patch

Review of attachment 8675799 [details] [diff] [review]:
-----------------------------------------------------------------

Yeah, this rushuffle looks fine. I agree that we don't particularly want to drill a JSContext all the way in. I was worried that we would be artifically delazifying too much, but it looks like we're actually using it beforehand, also.

Comment 6

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/dae20f86bbaa

Comment 7

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/dae20f86bbaa