695290 - Assertion failure: frame.stackDepth() == opinfo->stackDepth, at methodjit/Compiler.cpp:1672

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Christian Holler (:decoder)]

The following test crashes on mozilla-central revision cf5da681d577 (options -n -m -a):


options("typeinfer");
gczeal(2);
var g = newGlobal('new-compartment');
gcparam("maxBytes", 22000);
let array = g.eval("new Array(1,2,3)");


I'm not sure if this testcase is perfectly valid because the options call disables certain options again (but you cannot directly start with -n only). The larger testcase ran inside a driver which deletes the option function entirely though so I suspect there is a way to trigger this without using options().

Comment 1

[Brian Hackett [Laid off!]]

Attachment: patch

OOM triggered by inlineCallHelper was not being propagated.

Comment 2

[David Anderson [:dvander] - inactive, e-mail if emergency]

Comment on attachment 570023 [details] [diff] [review]
patch

Review of attachment 570023 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/methodjit/Compiler.cpp
@@ +3518,4 @@
>      if (!cx->typeInferenceEnabled()) {
>          CompileStatus status = callArrayBuiltin(callImmArgc, callingNew);
>          if (status != Compile_InlineAbort)
> +            return false;

Wouldn't this return false if status == Compile_Okay?

Comment 3

[Brian Hackett [Laid off!]]

Yes, that should be returning true on Compile_Okay, good catch.

https://hg.mozilla.org/integration/mozilla-inbound/rev/1814bb5292a4

Comment 4

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/1814bb5292a4

Comment 5

[Christian Holler (:decoder)]

Automatically extracted testcase for this bug was committed:

https://hg.mozilla.org/mozilla-central/rev/efaf8960a929