1Password crashes during sync on devices using an Intel processor [Fixed in version 4.2.1]

centurypixel

Oops, sorry, I forgot to put the smiley face. Here: :)

@saad and @peri, no pun intended for you guys. Please don't misunderstand, but if you put yourselves in our shoes and hear the similar response for 7 months, I think you will start making fun and enjoying the ridiculousness for a bit, eh?

blugos

Enjoying the ridiculousness for a bit

Yes, but not any kind of bit! It would be an agile bit :p

centurypixel

Sir, you got it!, a "bit" 8-)

peri

No offense taken for the jokes @centurypixel. :) We understand that this is a frustrating issue for our users with Intel processors.

Unfortunately, as you've guessed, we don't have news on this issue. We've reached out to both Google and Intel, and our devs are still trying to come up with a workable solution.

Thanks for hanging in there! And thanks again! :lol:

Jaaaaaaaamie

Just got a new Hudl 2.

You know the rest.

taps watch

peri

Welcome to the forums @Jaaaaaaaamie. :) I believe I responded to your email earlier, but it looks like you found the forum post first.

I'm sorry that 1Password won't sync on your new Hudl 2 at the moment. Our devs are currently reaching out and trying to find a solution. We'll update this thread with news when we have it.

blugos

Ok, is it possible at least to know a little bit more on what is going on?
For example, are you compiling 1Password specifically for x86 along with arm, or are you just compiling for arm and relying on libhoudini to have your libraries executed on a x86 processor?
I suspect that you are relying on libhoudini because you are putting the problem in the hands of google and intel and apparently just waiting for them to solve the problem (probably a bug in libhoudini)…
If that is the case, why aren't you implementing native x86 code? Or at least rewriting the faulty library for x86?

If it is not the case, then what is the problem exactly for which Google and Intel are responsible for? What library? What other apps out there are issuing the same problem? How did you come to the conclusion that Google or Intel are faulty?
And in that case as well (that you are doing native x86 code), can't you use libhoudini with your working arm code until the x86 bug is solved?

We need (at least I do) to know more on that, because from the android x86 customer point of view, it seems that you are not doing much but wait on Google and Intel… 7 month that a major bug making the app useless is known, and nothing seems to have been done except faulting Google and Intel and waiting for them to solve "your" problem. I do not doubt of the fact that they might be indeed responsible of a bug preventing your code to work as-in on the x86 environment, so at least, point it to us what is that bug from Google or Intel! We could then contact them and urge them to fix it, adding our voices. Meanwhile, as I said previously, as a serious company, you certainly can code the faulty portion differently, find a workaround, a temporary fix to at least have your customers being able to use your product… You are not newbies: I have known 1Password at its beginning, on Mac only, then on iOS. Then you worked on Windows and now Android. I hope that you are really serious, be we cannot see that from where we are. We are just being told that it is a hard problem, and that we have to wait, that you are sorry; it does not look serious and it makes us, customers, feel unimportant, neglected. Please, try to put yourself in our shoes for a second. I jumped recently in that discussion and issue, but seeing that it has been a problem for 7 month already, and that there is no clear progress or feedback about the problem, I really am thinking of considering the competition, for all my devices.

Thank you.

eltel

@blugos Very well put and clearly you know a lot more about the specific software environment(s) than many of us. No doubt we'll just get another bland corporate speak "we're reaching out" response from @peri which quite frankly is just patronising to AgilebBits customers. I'll also flag this to a couple of web-based tech review and testing sites to see if they'll pick it up and get a response from AgileBits. It would be nice to get something from Dave Teare on this but he seems unwilling to comment and just let the foot soldiers take the flack. I'm trialling Dashlane on some devices now and it works very well.

blugos

Funny, I am trialling Lastpass on different devices as well, and it works fine so far…

Jaaaaaaaamie

Thanks for the reply, @peri . Sadly, given the timescale of this thread, I will not hold out too much hope for any 'reaching out' achieving anything. I won't flounce out the thread, stating I am going to start using other software, as I doubt that really bothers anyone.

p.s Great post by @blugos.

Giansauna

Any news about the Intel bug, AgileBits? I'm sitting here with my brand new Yoga 2 tablet and have to turn on my desktop for euch sensible information stored in 1password.
Should I change to another password manager software? Please give me an advice.

mverde

@blugos You have raised some excellent questions and I will do my best to provide answers to them here.

We are neither compiling 1Password for x86 nor relying on libhoudini to execute ARM libraries on x86. 1Password for Android is written entirely in Java and does not include any third-party native libraries. The only native code that is being executed on behalf of 1Password is code that is supplied by Android itself.

The crash happens on Intel devices during the sequences of decryption and encryption operations that occur as a result of sync. These routines make use of the cryptographic methods that are built into Android and are not supplied by third-party libraries.

The crash itself is the result of a segmentation fault that occurs in the system shared libraries. The backtraces from crash reports indicate that this occurs most often in libdvm.so, but also occassionally in libc.so. Given that these segmentation faults are only occurring on Intel devices, it would seem that there is some native code in the system libraries that is behaving differently on Intel processors than it does on ARM processors.

We are currently working with Intel to determine why this might be and how we can best resolve this issue. Although I can't give you a timeline as to when the issue will be resolved, I do hope that I've at least managed to answer some of your questions about it.

RichardPayne

Come on then @mverde, the latest release notes say this is fixed; how did you work around the fault? ;)

mverde

@RichardPayne The release notes for the latest beta do indeed say that we have implemented a fix for this issue :)

Resolving the issue was a bit of a trip down the rabbit hole. After a lot of digging, we were finally able to connect the segmentation faults that I mentioned previously to code that otherwise seemed valid. While we don't fully understand why this code caused problems on Intel devices and not ARM devices, we were able to rewrite the logic in such a way as to avoid the crash. After resolving this first crash, we discovered a second crash that also seemed to be specific to Intel devices. The second crash occurred during decryption of large attachments. While we generally limit the size of attachments in 1Password for Android, we ended up needing to be more restrictive on Intel devices and limiting them to 1MB in order to avoid the crash.

We are hoping that our beta testers can help us confirm this fix so that we can release it on Google Play as soon as possible. If anyone here is interested in helping us test the fix, we invite you to sign up for our Android beta newsletter which will provide the link to download the latest beta version to your device.

Jaaaaaaaamie

Just like to post and say syncing is indeed working on my Hudl 2 (Intel Atom).

Initially, I had done the 'spend 30/45 minutes syncing/crashing/signing in' to get myself running, but having faith, I nuked all my data and did a fresh sync of my 450 logins/notes etc.

I realise I did get lucky, as I bought my tablet 10 days ago. So only had to wait a short time for the issue to be fixed.

Regardless, thanks again for the fix.

I now retract my flounce :blush:

saad

Great to hear the Intel related crash is resolved on your device and you are able to sync again! Thanks for confirming this for us. :)

ewanb

Just tried the latest beta version on my Asus Memo Pad 7 and it still crashes whilst syncing. Happy to volunteer more information if it helps.

ewanb

Actually re-read the last update and checked and I had one attachment of 1.1MB. Removed that using my PC and then everything works fine on tablet. It might be worth warning the user when adding an attachment, or in the case of existing attachments not try to decrypt them on a device which will not be able to?

peri

Thanks for the info, @ewanb. I'm glad to hear that the crash you experienced was unrelated to the Intel crash (phew). :)

I'll pass your suggestion along! Let us know if you need anything else!

mqueue

I installed Version 4.2.1b3 (20150416) (10053) of the beta on my Asus MeMO Pad 7 (ME176CX-A1-RD, Model K013, Android 4.4.2) and 1Password no longer crashes. All items sync'd without issue.

peri

That's great to hear! Thanks for letting us know! :)

blugos

Back from vacation, and very glad to see movement! I installed the latest beta, and it does work! Thank you for finally listening, and for the more technical feedback about what was going on. I was about to totally switch to competition Lastpass having interesting features, but ugly design, but now I think I will stick with 1password.

saad

On behalf of mverde, you're very welcome. :) Glad to hear sync is now working on your Intel based device!

The fix for this crash has also been released on the Play Store.

Charles Butcher

Thank you, AgileBits folks. It took a long time, and we did get a bit frustrated, but it was obviously a difficult one and you nailed it in the end.

saad

Thanks Charles! We are very happy we were able to resolve this one! :+1:

centurypixel

Yes, it took a very long time, but the fix works great. Thank you guys!

peri

@centurypixel. Thanks for the update! I'm glad to hear all is working well for you now. :)

Cheers!