707911 - Rewrite StackTraceToString to use StringBuffer

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Comment 1

[Tom S. (please needinfo tschuster)]

Attachment: v1

I think this works around the problem in the old version of this.

So before I try to add name, line and file for a new StackFrame, I first try to reserve that space, this way we shouldn't OOM while adding either name, line or file and all entries should be correct.

I also kept around the length limit.

Comment 2

[Brian Hackett [Laid off!]]

Comment on attachment 624124 [details] [diff] [review]
v1

Review of attachment 624124 [details] [diff] [review]:
-----------------------------------------------------------------

Nice cleanup.

::: js/src/jit-test/tests/basic/bug707911.js
@@ +12,5 @@
> +	/* oom should not cause abort, but instead just shorten the stacktrace */
> +} catch (e) {
> +	/* we assume at least one entry make it into the stack */
> +	assertEq(e.stack.indexOf('a@') >= 0, true); 
> +}

Without a value for -A this test won't trigger an OOM while constructing the stack trace, it will hit recursion limits long before.  Either fix the test or remove it.

Comment 3

[Tom S. (please needinfo tschuster)]

http://hg.mozilla.org/integration/mozilla-inbound/rev/20b27f266d73

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/20b27f266d73