708261 - findReferences crashes in shell js/tests (js1_8_5/extensions/findReferences-01.js et al)

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jason Orendorff [:jorendorff]]

REGRESSIONS
    js1_5/Regress/regress-96526-003.js
    js1_8/extensions/regress-422269.js
    js1_8/genexps/regress-380237-01.js
    js1_8_5/extensions/findReferences-01.js
    js1_8_5/extensions/findReferences-02.js
    js1_8_5/extensions/findReferences-03.js
    js1_8_5/extensions/findReferences-04.js

Comment 1

[Jason Orendorff [:jorendorff]]

findReferences always crashes now.

The first bad revision is:
changeset:   81246:ff51ddfdf5d1
parent:      77573:44ef245b8706
user:        Brian Hackett <[e-mail]>
date:        Wed Sep 28 15:04:55 2011 -0700
summary:     Remove shape numbers and Shape::slotSpan, factor Shape getter/setter into BaseShape, bug 684505.

Comment 2

[Jason Orendorff [:jorendorff]]

Attachment: v1

It was previously possible to get output containing the property id for getters and setters, but it seems this is no longer feasible, since the graph is like this:

  shape -> propid
  shape -> base -> getter

I think multiple properties with different names can share 'base'.

Comment 3

[Jason Orendorff [:jorendorff]]

The above patch does not solve the problems with
    js1_5/Regress/regress-96526-003.js
    js1_8/extensions/regress-422269.js
    js1_8/genexps/regress-380237-01.js
all of which seem to be older issues, and perhaps only happen on Mac with the new compiler.

Comment 7

[Jim Blandy :jimb]

So is the bug here that MarkChildren is passing PrintPropertyGetterOrSetter, which expects the debugArg to be a Shape *, but supplying a BaseShape * instead?

Comment 8

[Jim Blandy :jimb]

findReferences should just report the graph edges that the GC gives it; it needn't jump through hoops to try to get getter/setter names. So getting 'shape; base; getter' instead of 'shape; foo getter' is totally fine.

Comment 9

[Jason Orendorff [:jorendorff]]

(In reply to Jim Blandy :jimb from comment #7)
> So is the bug here that MarkChildren is passing PrintPropertyGetterOrSetter,
> which expects the debugArg to be a Shape *, but supplying a BaseShape *
> instead?

Yes. However, since the BaseShape doesn't have the information PrintPropertyGetterOrSetter used to print, there's no point using a debugPrinter function there at all.

(In reply to Jim Blandy :jimb from comment #8)
> findReferences should just report the graph edges that the GC gives it; it
> needn't jump through hoops to try to get getter/setter names.

Great.

Comment 10

[Jim Blandy :jimb]

Comment on attachment 579876 [details] [diff] [review]
v1

Review of attachment 579876 [details] [diff] [review]:
-----------------------------------------------------------------

Splendid, splendid. Thanks for sorting this out!

Comment 11

[Jim Blandy :jimb]

https://hg.mozilla.org/integration/mozilla-inbound/rev/e7785a537e5d

Comment 12

[Jason Orendorff [:jorendorff]]

Changing the description to be more specific. The rest of the failing tests will have to get their own bugs. I made one for the "too much recursion" stuff. In fact I made two, but bug 708862 is the better report.

Comment 13

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/e7785a537e5d