Closed Bug 1140287 Opened 10 years ago Closed 9 years ago

v36.0 slow to start up

Categories

(Core :: General, defect)

36 Branch
x86
Windows Vista
defect
Not set
normal

Tracking

()

RESOLVED INCOMPLETE
Tracking Status
firefox36 - wontfix
firefox37 - wontfix
firefox38 - wontfix
firefox39 - wontfix
firefox-esr31 --- wontfix

People

(Reporter: nigelcoxadi, Unassigned)

Details

(Keywords: perf, regression)

User Agent: Mozilla/5.0 (Windows NT 6.0; rv:35.0) Gecko/20100101 Firefox/35.0 Build ID: 20150122214805 Steps to reproduce: After upgrading to v36, Firefox takes up to a minute to load. I have reverted to v35.0.1 and will stay there until someone acknowledges this issue and states it is fixed. Actual results: As above, very slow to load. This is on Win7 laptop, and on an XP desktop. others have this issue, see the Community thread: https://support.mozilla.org/en-US/questions/1048872#answer-699778 Expected results: It should load a lot quicker, like v35.0.1 Should have received an official reply on the community thread, at least to show someone is paying attention. If this is fixed, it should show in the list of fixes for v36.0.1. if not fixed, it should show in the list of unresolved for v36.0.1
I have same issue and I would like to give more details: First I have installed FF v 36.0, I have Windows 7, i7 CPU, 8gb RAM and SSD drive It takes 60-80 sec for Firefox to start also I have noticed that when shutdown it will take about same time for process to disappear. I have tried following: - Uninstalling old version, - Deleting manually folders in Program Files - Deleting profiles both in AppData/Local and AppData/Roaming - Installing new version - Disabling all the plug-ins and extensions - I have tried safe mode. - And even installing new graphic card drivers And nothing really helped; I also have "Norton Internet Security". I have done the suggested regarding the Norton Firewall, and I have restarted comp, although it seemed that it helped at first start time was shorten to 3 sec. But then I have noticed that every subsequent turn on/off will make FF starting slower 60-90 sec. I have remembered that same behavior happened before the Norton Firewall tweaks but I was not paying attention. One more thing I have noticed when I get this slow starts CPU fan will immediately start working in higher speed. (even though CPU sometimes does not get over 3%). Then I have installed “about:startup” plug-in, and I've disabled all the extensions and plug-ins except the one I have just mentioned. Times in the list are in descending order or last measurement is first, times are in milliseconds. As you can see at the beginning time ranges to 50-60 seconds after each FF restart. Then I have restarted Windows time dropped to 3-4 sec for the first two FF start, 3rd one was again slowed down to 61sec. Then I’ve remembered I have additional component called “Trusteer Rapport” I have uninstalled that one and time dropped from 61s to 14-20 sec, again I have done the Win restart and 1st start of FF dropped to 2.5 sec, subsequent start was again in the range from 13 to 30 sec now. My question is what is so significant about subsequent start of FF ? firstLoadURI 33815.06 13951 13706 2nd time after restart 2529 1st after restart 15325 14641 14297 19982 After Trusteer Rapport uninstall 61834 3rd after restart 3340 2nd time after restart 4370 1sr after restart 57445 60357 59894 55924 55912 56622 50682 My bank requires me to have Trusteer installed. I have also installed update 36.0.1, the issue persist on Windows 7. And it seems that leg is getting even slower.
This problem has returned with FF 36.0.3 and the security update put on this morning (UK time). Again, it is taking some 60 seconds from clicking a link in an email to the page opening. [Win 7, SSD C: drive]. I have Trusteer and Norton and will not take them off. This problem appears to have been previously solved by one of Trusteer or Norton. Why is it that Mozilla can't learn from that and not rush out so-called security fixes that then don't work with related users' security software? If there isn't a properly considered response from Mozilla I shall return to FF 35.0.1 Just to create an account and click the authorization link in the email from Bugzilla takes some 60 seconds! Not your finest hour.
I can confirm this problem has returned with FF 36.0.3 in my case it takes around 80 sec to load.
[Tracking Requested - why for this release]: We should track this, and make sure it isn't still happening with 36.0.4 and 37. Kyle, I'm looping you in on this in case it is related to the fixes to 36 from the last few days.
Status: UNCONFIRMED → NEW
Component: Untriaged → Security
Ever confirmed: true
Flags: needinfo?(khuey)
Product: Firefox → Core
Florin can your team test to see if this affects 36.0.4 and other channels? Thank you!
Flags: needinfo?(florin.mezei)
I can confirm this problem has affected FF v36.0.4 as well. Loading time since the click on icon: (About:startup extension) 57.394sec 38.134sec 52.018sec 61.127sec
When it first occurred on 36.0.1 I went back to FF 35.0.1 which worked fine, just as it had been doing before 36 arrived. Trusteer sent me a link to install their latest and greatest - hasn't solved the problem for me with 36.0.3
My level of Trusteer is Version: Emerald Build 1404.75
Trusteer Version: Emerald Build 1404.75 Norton Internet Security Version 21.7.0.11 Windows 7 Ultimate
I don't know that this has anything to do with security. This has been reported as a regression in Firefox 36.0. We should try to get a regression window to identify the source of the issue.
It's extremely unlikely that the security issues we fixed in 36.0.3 and 36.0.4 caused a significant performance regression. This was likely introduced earlier, or comes from some strange interaction with third party software (or both). Unfortunately I don't think I can be of much help here.
Flags: needinfo?(khuey)
Tracking 37+ as we have multiple reports. I have marked 36 as wontfix as we're not going to take any more fixes for 36 before 37 is released.
Moving to Core::General for now as I doubt that this is security related. dmajor - Without more information, I'm not sure who to ask for help in moving this bug forward. Thought I would start with you to see if you have any ideas for next steps to get some more useful debugging info?
Component: Security → General
Flags: needinfo?(dmajor)
In these comments and in the linked forums I see a lot of mentions of antivirus programs. It sounds like AVs are spending a lot of time scanning Firefox when it starts. I can't say for certain why this has gotten worse with 36; maybe we're doing something that the AVs consider suspicious and prompting a deep scan. Here are some wild guesses off the top of my head: - We're updating a lot and throwing off signatures - libxul is huge now that we merged a few DLLs - we removed a bunch of JS export APIs that some AVs like to hook - we're opening ports on startup for SSDP Again, these are all guesses. We can't know for sure without talking to the AV vendors. Can we get a hold of anyone at Norton or Trusteer to help?
Flags: needinfo?(dmajor)
(In reply to :dmajor (semi-away, use needinfo) from comment #15) > Again, these are all guesses. We can't know for sure without talking to the > AV vendors. Can we get a hold of anyone at Norton or Trusteer to help? ni myself as a reminder to follow up on this tomorrow.
Flags: needinfo?(lmandel)
From this morning I do not have the issue anymore. Now I am very confuse as version of everything stayed the same: FF Version 36.0.4 (start time 1.7 s to 2.8 s) Trusteer Version: Emerald Build 1404.75 Norton Version: 21.7.0.11 !?
I reviewed Telemetry data from 35.x and 36.x with Vladan today and it doesn't show any discernible change in start-up times from the release population. Dave - Can you pull start-up time information from FHR to try and confirm that there is no significant increase in 36.x from 35.x?
Flags: needinfo?(lmandel) → needinfo?(dzeber)
We have done some testing yesterday on two separate machines running Windows 7 x64 and Windows Vista x64 but we were unable to reproduce the issue with: 36.0, 36.0.4, 37.0b7, 38.0a2, 39.0a1.
Flags: needinfo?(florin.mezei)
I've been away for a few days. Installed FF36.0.4 this afternoon and startup and executing an email link are working well! Surprising, as apart from installing 36.0.4 I'm not aware of changing anything else. Trusteer is Version: Emerald Build 1404.75 and Norton is 27.7.0.11!
Based on the latest information that we have, this bug does not need to block the Firefox 37 release. Setting ni on myself again to speak with IBM about Trusteer.
Flags: needinfo?(lmandel)
I'm marking this bug as invalid but I am still going to follow up so that we better understand how the combination of Firefox and Trusteer may have caused the slow start-up.
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → INVALID
This is disappointing that you are marking this as both resolved and invalid when it is neither. You seem to be concentrating on Trusteer, yet if you re-read the bug, and refer to the community thread, you will see that I am not running Trusteer, neither am I running Norton. So the issue is not specific to them, but could be related to the way antivirus systems work, as suggested earlier. From a user viewpoint, this problem started with v36. So something YOU changed needs revisiting. Just because you cannot find the problem, does not mean it is not there.
Status: RESOLVED → REOPENED
Resolution: INVALID → ---
(In reply to Nigel_Cox from comment #23) > This is disappointing that you are marking this as both resolved and invalid > when it is neither. You seem to be concentrating on Trusteer, yet if you > re-read the bug, and refer to the community thread, you will see that I am > not running Trusteer, neither am I running Norton. So the issue is not > specific to them, but could be related to the way antivirus systems work, as > suggested earlier. My mistake. Thank you for reopening. I was going off of the comment in the support forum that stated that the problem was fixed "Update: Now installed 36.0.1 on the XP desktop, seems to be working fine there too. so this issue seems to be fixed." along with data showing that start-up time has not noticeably increased for the Firefox population as a whole in 36.x. I have contacted Trusteer to try to understand what will trigger scanning or other activities that may impact start-up time. My expectation is that the results will be generally applicable to AV software. > From a user viewpoint, this problem started with v36. So something YOU > changed needs revisiting. Just because you cannot find the problem, does > not mean it is not there. That's not always the case. In some cases other products need to update their whitelists or have a bug that is exposed in a new version. As a concrete example, there was a product that had custom logic for Firefox 3.x. Unfortunately, once the version of Firefox hit 30, their detection algorithm flagged it as 3.x and applied the wrong logic. (In reply to Marko from comment #1) > And nothing really helped; I also have "Norton Internet Security". > I have done the suggested regarding the Norton Firewall, and I have > restarted comp, although it seemed that it helped at first start time was > shorten to 3 sec. > > But then I have noticed that every subsequent turn on/off will make FF > starting slower 60-90 sec. > I have remembered that same behavior happened before the Norton Firewall > tweaks but I was not paying attention. > One more thing I have noticed when I get this slow starts CPU fan will > immediately start working in higher speed. (even though CPU sometimes does > not get over 3%). This symptoms sound like a low physical memory condition. What is your RAM usage when you have slow start-up? If you're over your RAM limit your machine may be paging to disk which will slow down the entire machine.
I did not see any significant difference between 35.x and 36.x in FHR data either. I looked at times for all three startup phases across major hardware breakdowns (as recorded in FHR) for the population of release channel users on Windows.
Flags: needinfo?(dzeber)
Although I know that this issue is not Trusteer specific, I did speak with Trusteer today. Here is something to try for anyone who experiences this issue and does have Trusteer installed from Trusteer support: "Here are the instructions to stop Rapport to verify whether the slowness issue is indeed with Rapport. To stop and start Rapport: PC users - Select Start > Programs > Trusteer Endpoint Protection > Stop / Start Trusteer Endpoint Protection Mac users - Select Apple Menu > System Preferences > Other > Trusteer Endpoint Protection > Stop / Start Rapport Please note that this is not a permanent solution, and while Rapport is deactivated you are not protected. If you indeed determine that the issue is caused by Rapport, we would appreciate if you can forward us the feedback as well as direct users to contact us so we may troubleshoot the issue."
Version 37.0.1 issue is back. Start time 82 sec.
This really sounds like the AVs are doing extra scanning on FF builds when they are new and not yet recognized.
Just installed FF 37.0.2 and FF is slow to start again. I've gone thru the procedure Trusteer told me to follow AND restarted and still FF is slow to start and slow to fire up when a link is clicked in another program. This is just so frustrating. When will Mozilla work out how to update FireFox without causing all this waste of time and energy restarting systems and raising error reports with Mozilla, Trusteer and Norton?
(In reply to Lawrence Mandel [:lmandel] (use needinfo) from comment #24) > (In reply to Marko from comment #1) > > And nothing really helped; I also have "Norton Internet Security". > > I have done the suggested regarding the Norton Firewall, and I have > > restarted comp, although it seemed that it helped at first start time was > > shorten to 3 sec. > > > > But then I have noticed that every subsequent turn on/off will make FF > > starting slower 60-90 sec. > > I have remembered that same behavior happened before the Norton Firewall > > tweaks but I was not paying attention. > > One more thing I have noticed when I get this slow starts CPU fan will > > immediately start working in higher speed. (even though CPU sometimes does > > not get over 3%). > > This symptoms sound like a low physical memory condition. What is your RAM > usage when you have slow start-up? If you're over your RAM limit your > machine may be paging to disk which will slow down the entire machine. Marko?
Flags: needinfo?(marko77pet)
Keywords: regression
Perhaps last part of comment 24 was directed to Nigel. Without more information this isn't actionable.
Status: REOPENED → RESOLVED
Closed: 10 years ago9 years ago
Flags: needinfo?(marko77pet)
Resolution: --- → INCOMPLETE
> Without more information this isn't actionable. Likewise, assuming the problem is gone for most/some of you, it would be nice to hear from you about how and why it changed behavior.
You need to log in before you can comment on or make changes to this bug.