842623 - Gerrit CPU hog - unresponsive script warning

Status: RESOLVED WORKSFORME

(Core :: JavaScript Engine, defect)

Description

[Andre Klapper]

Forwarding from https://bugzilla.wikimedia.org/show_bug.cgi?id=35860 

Original report:
"Gerrit is a browser memory hog and/or makes my browser hang when "Diff all
side-by-side" of https://gerrit.wikimedia.org/r/#change,4631. Seems to be
caused by tests/messageindexdata.ser. I think Gerrit probably shouldn't just
open these kind of files."


What I get with Firefox 18.0.2:

A script on this page may be busy, or it may have stopped responding. You can stop the script now, or you can continue to see if the script will complete.
<andre__> Script: https://gerrit.wikimedia.org/r/gerrit_ui/CF42DB02C9A85777BE331C90AA0419DE.cache.html:4386


<^demon> Works fine for me in safari and chrome.
<^demon> But yeah, can definitely replicate in firefox.
<^demon> 8,388,608 B (10.28%) ── string(length=4181450, "a:49286:{s:19:&quot;8:filereado...") looks fairly suspicious.

Comment 1

[Andre Klapper]

Attachment: about:memory?verbose output

Comment 2

[(mostly gone) XtC4UaLL [:xtc4uall]]

Confirmed with Mozilla/5.0 (Windows NT 6.1; WOW64; rv:22.0) Gecko/20130224 Firefox/22.0 ID:20130224031053 CSet: 195e706140d1.

FWIW, Chrome also gives an "unresponsive Site" Dialog (with strangely only one Tab opened after hitting "All Side-by-Side") and its Content View Responsiveness is blocked.

Comment 3

[(mostly gone) XtC4UaLL [:xtc4uall]]

Still an Issue against Mozilla/5.0 (Windows NT 6.1; WOW64; rv:26.0) Gecko/20100101 Firefox/26.0 ID:20130810030206 CSet: c5946a8bcd5b

Comment 4

[Wayne Mery (:wsmwk)]

I don't see this when using 26.0a1 on wins7

(In reply to Andre Klapper from comment #0)
> Forwarding from https://bugzilla.wikimedia.org/show_bug.cgi?id=35860 
> 
> Original report:
> "Gerrit is a browser memory hog and/or makes my browser hang when "Diff all
> side-by-side" of https://gerrit.wikimedia.org/r/#change,4631. Seems to be
> caused by tests/messageindexdata.ser. I think Gerrit probably shouldn't just
> open these kind of files."

how is this a confirmed firefox issue?  Esp if it also happens in chrome?

Comment 5

[Andre Klapper]

Wasn't aware that Chrome also has issues when I created this.

Comment 6

[banakon]

watching a big foto on wikimedia, total crash
https://crash-stats.mozilla.com/report/index/cf9aa02d-982e-4e33-845d-aaa3b2131116

Comment 7

[Andre Klapper]

banakon: Unrelated. This report is about gerrit.wikimedia.org and not about images.

Comment 8

[Evan]

I am also seeing this issue on the Gerrit installation used by the Wireshark project (code.wireshark.org/review). Firefox 27.0.1 x86_64 on Ubuntu. Have not seen this with chrome, but don't use it nearly as much.

Comment 9

[Andre Klapper]

Removing needinfo as per comment 5 and 8.

Comment 10

[Evan]

This is strictly a guess, but after seeing this a couple of different times for Wireshark and doing a bit of digging, it seems that the time/memory taken is proportional to the size of the file being diffed (not the size of the diff). For normal files it doesn't really show up, but when the file is many lines then the diff page can take minutes to load.

For example, this page seems to trigger it reliably for me:
https://code.wireshark.org/review/#/c/696/2/epan/enterprise-numbers,cm

The underlying file is ~170k lines, while the diff is only 82 lines. The page still takes minutes to load. Does this take ages to load in other browsers as well? If so, this is a Gerrit bug (or a Codemirror bug, since IIRC that's what Gerrit uses for diffs) and not a Firefox bug at all.

Comment 11

[Wayne Mery (:wsmwk)]

(https://gerrit.wikimedia.org/r/#change,4631 + "All Side-by-Side" is too broad a testcase - it opens about 10 URLs)

Problem likely related to the fact that https://gerrit.wikimedia.org/r/#/c/4631/1/tests/messageindexdata.ser is a huge, long, 1 line blob (several MB?).  Therefore https://gerrit.wikimedia.org/r/#/c/4631/1/tests/messageindexdata.ser sucks CPU. Also in IE and chrome.

Comment 12

[mike bayer]

I am getting *killed* by this bug constantly on gerrit.  I've been having sporadic "unresponsive script" errors on other sites as well that are otherwise working for everyone else, so since then I've created a brand new profile - the errors went away for a week or so, now they're back.  If I make another brand new profile, same issue.  If I run in "safe mode" - same crash.  It's not the plugins.   Running under Safari, page comes right up.  

The link in question that is impossible for me to open in firefox at this moment is: https://review.openstack.org/#/c/106367/5/nova/tests/db/test_db_api.py

Comment 13

[Christian Höltje]

Does turning off "Use hardware acceleration when available" help?

Comment 14

[Wayne Mery (:wsmwk)]

Mike, Andre? 

(In reply to Christian Höltje from comment #13)
> Does turning off "Use hardware acceleration when available" help?

Comment 15

[mike bayer]

from my end I haven't had this problem in a long time, neither with the openstack gerrit nor with the one I run on sqlalchemy.org, on my end this is now WORKSFORME - given that both firefox and gerrit have seen many releases since my last comment.   I'll ping back here if I start seeing it again.