1089932 - Ignore replace_malloc allocation function pointers

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Steve Fink [:sfink] [:s:]]

glandium is running into problems where he sees thousands of analysis failures. They all boil down to something like:

    PR_Calloc
    calloc
    replace_calloc
    FieldCall: malloc_table_t.calloc

(ie, all of the allocation functions go through a function pointer that we don't recognize.)

Oddly, this failure is intermittent. I believe that is because normally replace_malloc is loaded via LD_PRELOAD, so the analysis doesn't see it. But something compiles it in directly, possibly just a test file. I am guessing that that something has the same name as something else, and it's a race to see which one gets compiled last (the later ones will override.)

glandium's patch changes compilation ordering, which would be why it would start failing reliably for him.

I'm a little curious why we have the same filename compiled multiple times in different ways, but I guess that's not really all that uncommon.

Comment 1

[Steve Fink [:sfink] [:s:]]

Attachment: Ignore replace_malloc allocation function pointers

The fix is straightforward even if the problem is not.

Comment 2

[Mike Hommey [:glandium]]

The something that links replace-malloc directly is memory/replace/logalloc/replay.

Comment 3

[Mike Hommey [:glandium]]

(In reply to Mike Hommey [:glandium] from comment #2)
> The something that links replace-malloc directly is
> memory/replace/logalloc/replay.

... but it only links the _impl part, not the part that *does* use malloc_table_t...

Comment 4

[Steve Fink [:sfink] [:s:]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/5422ee0c1808

Comment 5

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/5422ee0c1808