Closed Bug 756216 Opened 12 years ago Closed 5 years ago

LSP Causes Firefox to Load pages slowly

Categories

(Firefox :: General, defect)

Product:
Firefox
Component:
General
Platform:
x86
All
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED INCOMPLETE

People

(Reporter: coverton, Unassigned)

Details

Reproduction steps (I've reproduced this on 32-bit XP SP3 and Win 7):

(1) Install Firefox
(2) Install CyberPatrol Online Protection Pro.
(3) Try to load a web page in Firefox (the page I used for testing was www.slashdot.org).

Firefox spins for up to 3 minutes before finally loading the page.

In some cases, pages load more quickly.  But this can be easily reproduced by trying 2 or 3 different sites.  The issue seems to come and go from site to site, but can be easily reproduced.

Unhooking the LSP for Online Protection Pro resolves this issue.  However, this is not an LSP crash, nor a Firefox crash.  The issue is that the LSP is doing something that is causing Firefox to spin.  The same LSP has no issues with other browsers.

Also, I _think_ this exact same LSP worked properly with Firefox 9.  So this issue may have been introduced some time since that version.  At the very least, it's gotten worse in versions greater than 9.0.

If need a copy of the Online Protection software to test with, please let me know.  I'm a developer at CyberPatrol and we've been trying to track this issue down for some time.  It doesn't appear to be anything in our LSP code, but I'd like to work with someone doing Firefox development to get to the bottom of it.
Hi! Thanks for reporting, usually it's us jumping through hoops to get in contact with developers. :)

Do you have any stacks / profiling data about where things are going wrong?

Is this this consistently reproducible with your steps above?

CCing a few folks.
Status: UNCONFIRMED → NEW
Ever confirmed: true
I'd like to see if I can reproduce it, is there a trial version I can use to try to reproduce it?
From what I can tell, Firefox goes through more or less the standard calls, and then gets to a point where it seems to spin trying to "select" on a specific socket over and over.

I'm not in a place where I can generate a product license at the moment, but I'm happy to provide a link to a full license key and install instructions first thing tomorrow.  Is there an email address I can send the information to?  I'd rather not post the product key as part of a public bug discussion.
Forgot to answer Justin's questions:

(1) Thanks to you guys for getting to this so quickly.  I've had much less responsiveness from developers of other products in the past.

(2) Yes.  I can consistently reproduce it with Firefox 12.0 and our product on XP and Win7.  I suspect it would also reproduce on Vista, but I haven't tried it so far.  It actually seems that it doesn't matter which order you perform the first two steps in.  Regardless of which software you install first, the results are the same.
(In reply to coverton from comment #3)

> I'm not in a place where I can generate a product license at the moment, but
> I'm happy to provide a link to a full license key and install instructions
> first thing tomorrow.  Is there an email address I can send the information
> to?  I'd rather not post the product key as part of a public bug discussion.

Sure. I can confirm Brian Bondy works for Mozilla, so feel free to email him (netzen@gmail.com) directly. If you'd prefer an "official" address, bbondy(at)mozilla.com should work as well.
Thanks.  I've sent Brian a license via email, and instructions for installing the software.  I'm happy to work with you guys to track down and fix this issue, so please keep me posted if you need help from my side.
Thanks! I received the license.  I'll be away on the weekend and there is a holiday in Canada on Monday.  I'll try to reproduce this next week though and get back to you with my findings.
I can reproduce with Firefox 12, but I cannot reproduce with Firefox Beta, Aurora, and Nightly.

For Firefox Release I can see using wireshark that:
The HTTP response header shows a content length of 99,439 but the whole response is not received (only 78,014 bytes of the content). 
So the browser sits waiting for more.

Can you confirm that you cannot reproduce with Aurora and Nightly?
Can you think of any reason why it would work with Beta, Aurora, and Nightly but not Firefox 12?
Maybe something changed in Beta, Aurora, and Nightly and it will work on the next release?
Does the install path matter? (I don't think so)
What page is the HTTP request/response for?
slashdot.org
I actually haven't tested with Beta or Aurora so far.  I will run those tests and get back to you.  I can't think of any reason (other than a bug fix, or other code change) that might cause a difference.

Also, I've never tried using a different install path.  But I don't suspect it will make any difference at all.
I've tested this with Firefox Beta (13.0b4) and I can confirm that it doesn't reproduce with this build.  I am unable to test the most recent nightly build (15.0a1).  I've downloaded it multiple times on different machines, and I always get "The file is corrupt" error from Windows when trying to run the installer.
OK maybe the best thing to do is to wait for Beta -> Release on June 5th and to re-test at that point.  Does that sound reasonable?
I think so.  Are you saying that the current Beta will go "live" on June 5?
That's correct, should go live very close to, or on June 5th.
https://wiki.mozilla.org/Releases
(In reply to coverton from comment #14)
> I think so.  Are you saying that the current Beta will go "live" on June 5?

coverton, does this still reproduce?
Flags: needinfo?(coverton)
Status: NEW → RESOLVED
Closed: 5 years ago
Flags: needinfo?(coverton)
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.