Bug 232328 – Fails to execute JavaScript when soft stack limit is unlimited

• Component: Web Template Framework
• Status: RESOLVED FIXED

[Michael Catanzaro]

Reproducer:

$ ulimit -s unlimited
$ jsc
>>> 1 + 1
Exception:

I'm not sure why JSC cannot print the exception, but it's "maximum call stack size exceeded."

Problem is since r276695 "[WPE][GTK] More correct fixes for stack size issues on musl libc", WTF::StackBounds::currentThreadStackBoundsInternal now calls getrlimit() to get the soft stack limit. When the limit is unlimited, rlim_cur is -1 and the function is not prepared for that. I don't know how to fix it other than to fall back to some suitable default (8192 seems to be the default soft limit).

This is only a regression on Linux, but the Darwin code has the same bug too.

[Daniel Kolesa]

let's have it max out at 8M, realistically there is nothing better you can do i think

[Michael Catanzaro]

Actually I'm not certain it's really a regression, because there are preexisting calls to pthread_getattr_np() (or pthread_get_stacksize_np() on Darwin) that would likely have had the same problem.

(In reply to Daniel Kolesa from comment #1)
> let's have it max out at 8M, realistically there is nothing better you can
> do i think

Sounds good.

[Michael Catanzaro]

(In reply to Michael Catanzaro from comment #0)
> When the limit is unlimited,
> rlim_cur is -1 and the function is not prepared for that.

Actually, it is RLIM_INFINITY. Doesn't change the rest of the analysis, though.

[Michael Catanzaro]

Created attachment 442530 [details]
Patch

[Michael Catanzaro]

(In reply to Michael Catanzaro from comment #2)
> Actually I'm not certain it's really a regression, because there are
> preexisting calls to pthread_getattr_np() (or pthread_get_stacksize_np() on
> Darwin) that would likely have had the same problem.

I think it's OK because this returns the real stack size, not the limit. Also, it works fine in practice with no changes to this.

[Daniel Kolesa]

patch lgtm

[Radar WebKit Bug Importer]

<rdar://problem/84943302>

[Michael Catanzaro]

Ping reviewers

[Yusuke Suzuki]

Comment on attachment 442530 [details]
Patch

r=me

[EWS]

Committed r285187 (243816@main): <https://commits.webkit.org/243816@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 442530 [details].