410488 - 100% CPU during Weave sync, UI unresponsive

Status: RESOLVED DUPLICATE of bug 409742

(Cloud Services :: General, defect)

Description

[George K.]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b2) Gecko/2007121120 Firefox/3.0b2
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b2) Gecko/2007121120 Firefox/3.0b2

When performing a Weave synchronization (either automatic or triggered by clicking the "Sync Now") the Firefox UI becomes unresponsive and one CPU gets pegged at 100%.

Some time after this, I get a "Warning: Unresponsive script" message telling me the script "... may be busy, or it may have stopped responding."  The problem script has always been the \extensions\{340c2bbc-ce74-4362-90b5-7c26312808ef}\modules\xxtea.js file (rendered as "/extensions/%7b340c2bbc-ce74-4362-90b5-7c26312808ef%7D/modules/xxtea.js in the warning), and the only lines I have seen in the message are 157 and 161.

Some time after the warning has appeared (and apparently no matter whether I have dismissed the warning or not) the UI becomes functional again (stops blocking), though the CPU load will remain at 100% for several more minutes as the sync completes.

The lines cited in the "busy script" messages are the escCtrlCh() and unescCtrlCh() functions, and the Verbose log shows the process is within one of the two following blocks when this happens:

Either:

Service.Crypto	DEBUG	Encrypting data
Service.Crypto	DEBUG	Done encrypting data

Or:

Service.Crypto	DEBUG	Decrypting data
Service.Crypto	DEBUG	Done decrypting data

I have Weave 0.1.13 installed on 2 Windows XP machines, one running Firefox 3.0 Beta 2 and one running Minefield nightly builds (currently Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b3pre) Gecko/2008010104 Minefield/3.0b3pre) and the issue affects both.

Removing "Browsing History" from the sync process seems to help a little with the amount of time the UI spends blocked, presumably since there is less data to scan for characters that might need escaping or unescaping.

The only similar bug I found is Bug 409742, which may be related (both files implement XXTEA) but if that bug is really resolved in 0.1.12.11 then this is not a duplicate.

Reproducible: Always

Comment 1

[Dan Mills [:thunder]]

Hey Andy,

Thanks for the additional info.  I think we should roll this bug into bug 409742.  I'm surprised you're still getting the busy script dialog, but it's possible the behavior is platform specific, or maybe there's something else I overlooked.

I'll do some profiling shortly.