Solved Browser crashes on startup
-
I have salvaged my tabs.
I essentially did this from earlier in this topic: https://forum.vivaldi.net/post/573941
except I installed VMWare Workstation and Win10, and copied the Session data over the virtual network. It's easy to toggle the network on and off in VMWare.I found two tabs that were trying to submit forms, closed them, saved the session (watch for the process to exit inside the VM, if you're trying this, as it will take a while), and reopened the session with the latest version (5.2.2623.39)
-
@pafflick Like @dols wrote, Vivaldi works a little longer, but after a day of use, I had again a crash. It happens when I try to write in an inputfield in Contao-CMS (Modulsearchbar at top). This is reproducable, everytime i do this, Vivaldi crashes.
The good thing, the profil stil is startable, thats a little win
Vivaldi Version: 5.2.2623.39
-
@mikefmmedia OK, this is crazy. I tried different Websites with input, all crashed the Browser when I started typing. Now, after some restarts, problem disappeared, I can't reproduce it at the moment.
I think this looks like a browser internal storagelack, that fills up some storage until it exeed and then it crashes. The restarts at some point may have cleared this storage and problem is gone.
I bet this will work until some storage will exeed again. Let's see maybe in a view hours or tomorrow... -
@mikefmmedia
Hi, you could try to open the internal task manager (Tools menu), maybe a tab or extension leaks memory. It should stay open after a crash.Cheers, mib
-
Ppafflick marked this topic as a question on -
Ppafflick has marked this topic as solved on
-
I also had the Vivaldi-Snapshot version crash immediately on startup after last night (no issues before that). Thanks to the tip about right clicking the icon I have pinned to the taskbar I was able to successfully launch it via the "new tab" option. I found a single tab open from last night that wanted form data. I usually close these and start with the Vivaldi start page.
Closing the tab got things working again for me. An update should fix the issue permanently.
Vivaldi 5.3.2649.3 (Official Build) (64-bit)
Revision cd672b0ff6ae10a43c069974a6fa55c0bf1df7c1
OS Windows 10 Version 21H2 (Build 19044.1645) -
I've started using Sync on both of my desktops, different accounts for each, and I see that Sessions are at least eligible to be synced, according to vivaldi://sync-internals . Let's say that I encounter this issue again in the future and I want to restore from my sync (I understand that it's not intended to be used as a backup) , would upgrading the browser cause the synced data to not be backwards-compatible with the previous version, such that I would be unable to repair a broken tab?
-
@dols I'm not sure you can access tabs from the same instance of the browser that sync got them from. And then, the tabs themselves are not stored - just their position and state (pinned or stacked) and URL. To store everyone's synced tabs in toto would require an impossibly huge amount of storage.
And then, too, tabs are not sync'ed in real time, so if you have, say, 31 tabs open, only 29 of them might have already been saved in sync.
So I would not count on this as a strategy.
Last note: Sync is only available from one machine to another if they are on the same account.
-
@rskunath I can confirm that the snapshot update from today fixes the tab open when Vivaldi is closed, and then Vivaldi crashes on relaunch because of empty fields in the tab, issue for me.
Vivaldi 5.3.2652.3 (Official Build) (64-bit)
Revision 7cb3851260722b78a28e04b95e342a9932aecc9d
OS Windows 10 Version 21H2 (Build 19044.1645) -
This post is deleted! -
This issue does not seem to be actually resolved at all. When I updated to 6.2, a random tab opening (I think it was from NY Times), crashed the browser, and I could not get it to start again until I rolled all the way back to 5.2.
This is running on XFCE 4.18, Linux 6.5 kernel, Debian.
I just upgraded to Vivaldi 5.8.2978.36 stable to see if that would help, but didn't.
Note that private window mode opens fine.
This just happened again, so I'm going to have to give up on vivaldi because there is simply no way to recover from this, and I rely on working session management since I used a lot of tabs and windows, and spent a long time organizing tabs into tab groups etc.
The only way I could get vivaldi to even start the last time this happened a few days ago was to roll back to vivaldi 5.2 using the .deb package manual install, that did allow the session to sort of halfway load, but very damaged because the refresh feature was totally broken and the only way to actually get the pages to load (they just showed a white page though the tab icon was correct, and the hover over icon view mode was showing the page from somewhere or other) was to use ctrl + L to make navigation url box active, then to hit enter manually. This is obviously not viable long or even short term.
I'm sad because I had a lot of information in my tabs, and now I clearly will have to find another browser. Vivaldi has been fairly unstable for a long time, it would randomly just decide to not work, but a force killall vivaldi-bin and vivaldi restart always worked in the past.
There appears to be a host of things broken in the session management, but the thing that gets me is that the entire point of using the Blink rendering engine is that each tab is a process, and is supposed to be sandboxed, so a new tab should not be able to crash the entire browser in this way, that seems to be a fundamental error that either was brought back in 6.2 or earlier, or which was never really actually fixed.
I can as a crude hack reinstall 5.2, which will get me my tabs back.
I've tried all the command line startup options, disable extensions (never the problem), don't use hardware acceleration (Never the problem), and the suggestion to wipe my sessions is clearly pointless since this is happening on specific tab load events so it will just happen again until the actual source of the problem is found and corrected.
I just used the install 5.2 method again, which allows vivaldi to start, then looking for the tab in question, closing it, then reinstalling current vivaldi, but this appears to fundamentally corrupt the session data breaking page reload and refresh completely.
I was going to give this a few version updates to see if the core issue is corrected, but it's not, a random url will cause the crash event, forcing this sequence of recovery steps, leading to a largely unusable system.
I can see crashes etc, but this is clearly something much more core because a tab should not be able to crash a blink based rendering engine browser in the first place, the crash is supposed to be restricted to the tab itself, but that is clearly totally broken, as is the recovery mode, and there is no safe mode at all that I can find, something that lets you load the tabs in restricted use mode, but even that I think is not relevant since the very fact that a tab can just crash the entire setup makes vivaldi unable to be an actual working browser, which is a drag because I like vivaldi, but something is seriously wrong internally, but external to the actual blink engine itself, which means it's in the vivaldi wrapper itself, which I can't fix, and can't rely on anymore.
I wish there were some magic trick to let a clean recovery happen.
I used many windows, 4-7 usually, with hundreds of tabs, and I picked vivaldi specifically because it's tab and session management was really robust, but I think something has gone very wrong in the code, and since I am posting on what was originally a 5.2 version thread with the same bug, I think it's quite safe to say the bug is not actually fixed, or was briefly fixed, and got broken again due to some oversight in coding.
To reiterate, allowing a URL for whatever reason to take down the entire browser, and to then break restart, and having no way to override this in some way to allow at least for finding and closing the offending tab makes this browser more or less unfit for purpose as far as I can see.
I saved my sessions when this happened initially a few days ago ( had just upgraded my system, probably bringing vivaldi from 5.x to 6.2 or 6.4, which is when this issue started.
Clearly, if a solution is dumping the Default profile directory, then vivaldi isn't usable since it takes forever to setup those sessions and tabs etc.
Note that there are some kind of serious issues here, primary being that there does not appear to be anyway to force start in debugger mode to allow for filing a bug report, thus making this issue basically impossible to debug beyond noting that it exists and is happening and is most definitely not corrected.
Normally I would not necropost, but given the claim was this was fixed, and it's most definitely not fixed, I think it's appropriate to link these issues over time to serve as a reminder the cause is most likely not fixed at core, though it may have been worked around a bit short term.
If I were you guys, I'd do as first step adding in a real debugger mode so that at least you can start and capture all these beyond what you get on a CLI failed start:
[89861:89861:1027/160659.987965:ERROR:chrome_browser_cloud_management_controller.cc(162)] Cloud management controller initialization aborted as CBCM is not enabled.
[89861:89896:1027/160700.013009:ERROR:login_database.cc(836)] Password store database is too new, kCurrentVersionNumber=33, GetCompatibleVersionNumber=40
[89861:89896:1027/160700.013380:ERROR:login_database_async_helper.cc(54)] Could not create/open login database.
[89861:89881:1027/160700.631489:ERROR:quota_database.cc(939)] Could not open the quota database, resetting.
[89861:89861:1027/160701.067051:ERROR:service_worker_task_queue.cc(234)] DidStartWorkerFail mpognobbkildjkofajifpdfhcoklimli: 18
[89861:89861:1027/160701.418372:ERROR:CONSOLE(1)] "Uncaught (in promise) [object Object]", source: chrome-extension://mpognobbkildjkofajifpdfhcoklimli/background-common-bundle.js (1)
[89861:89882:1027/160703.161768:ERROR:quota_database.cc(939)] Could not open the quota database, resetting.
[1027/160705.054307:ERROR:elf_dynamic_array_reader.h(64)] tag not found -
I'm hoping the actual cause of this is fixed (that is, a loaded tab can no longer crash vivaldi, or make vivaldi start/load fail when the tab starts loading on restart) so I can return to vivaldi, which is why I saved my session, but given this how now happened 2x in a few days, I have to consider current vivaldi as off limits as a primary browser, which is extremely unfortunate.
I am not positive, but this time's offending tab may have been the mercury news, a newspaper website, and the last one was I am fairly sure the New York Times, but I wouldn't swear to it. So those may contain something, but again, it should not matter what a legitimate tab/url contains, it should load fine.
While disabling extensions did not make any difference at all in terms of restart after crash, I am running ublock origin, but nothing else. That however should make it less, not more, likely that nothing crashes the page since it's blocking a bunch of nasty ad garbage and js.
-
Create easy "BAT" files:
start G:\Vivaldi2\Application\vivaldi.exe
sleep 20
taskkill /f /t /im "vivaldi.exe"
sleep 6
start G:\Vivaldi2\Application\vivaldi.exe
Second startup is without problem
-
@h2sm You utter my sentiments exactly. Vivaldi's tab and session management were great, far exceeding any other browser's capabilities for users of many, many tabs like us. That is, until this crash-without-recovery issue arose.
The common suggestions to delete/rename one's profile etc. are completely beside the point, we well know that it's something in the saved session data that's causing the crash, but the saved session data itself is far more important than the choice of browser. Unrecoverable sessions are absolutely not acceptable.
Fortunately I (currently) have a feasible workaround of starting Vivaldi with a private window, then opening a new window, which happens to open the saved session as well (in yet another window). So clearly nothing in that saved session and its tabs is truly preventing Vivaldi from launching, it must be a really dumb error causing those startup crashes, some uncaught and unhandled exception, Programming 101 stuff.
