[Webkit-unassigned] [Bug 171630] New: [JSC] Test stringimpl-to-jsstring-on-large-strings-2 uses a lot of RAM

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=171630

            Bug ID: 171630
           Summary: [JSC] Test stringimpl-to-jsstring-on-large-strings-2
                    uses a lot of RAM
           Product: WebKit
           Version: WebKit Nightly Build
          Hardware: Unspecified
                OS: Unspecified
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: JavaScriptCore
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: clopez at igalia.com
                CC: benjamin at webkit.org, mcatanzaro at igalia.com,
                    ossy at webkit.org, sbarati at apple.com

The JSC test stringimpl-to-jsstring-on-large-strings-2.js needs large amounts of RAM to run (between 3GB and 5GB on the GTK+ port)

I detected this because I just set a 16GB of RAM limit for the GTK+ release test bot (in order to detect things like this one), and this test crashed after doing that:

>From https://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release%20%28Tests%29/builds/902/steps/jscore-test/logs/stdio:
jsc-layout-tests.yaml/js/script-tests/stringimpl-to-jsstring-on-large-strings-2.js.layout-no-cjit: Killed

On the bot this was reported:
[ 1230.024807] Memory cgroup out of memory: Kill process 10486:#40003 (jsc) score 150 or sacrifice child
[ 1230.025888] Killed process 10486:#40003 (jsc) total-vm:5009940kB, anon-rss:2517060kB, file-rss:2568kB


Trying to guess if this was a WebKit bug or not I created a simple test page that just loads this test on the browser:

https://people.igalia.com/clopez/wkbug/largestrings/stringimpl-to-jsstring-on-large-strings-2.html


And the results I got are:

 - WebKit uses more than 3GB of RAM for loading that page. And the memory pressure handler doesn't do any magic here. I ran the minibrowser inside a linux cgroup memory limited to 3GB and the kernel killed the WebProcess because OOM

[479328.854217] [ pid ]   uid  tgid total_vm      rss nr_ptes nr_pmds swapents oom_score_adj name
[479328.854725] [10576]  1000 10576   547093    14894     192       5     2180             0 MiniBrowser
[479328.854733] [10596]  1000 10596   512439    12800     183       5     4157             0 WebKitNetworkPr
[479328.854737] [10599]  1000 10599  1374977   487480    1721       9   308803             0 WebKitWebProces
[479328.854749] Memory cgroup out of memory: Kill process 10599 (WebKitWebProces) score 1014 or sacrifice child

 - Chrome uses just 100MB of RAM (with a peak of 130MB of RAM) to load this very same page.

So, I think something is wrong here 

Looking at the bug that added this test, this was already reported by Ossy in its day: https://bugs.webkit.org/show_bug.cgi?id=158793#c6

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20170503/3f48ab94/attachment-0001.html>