Crash with encryption - meta-issue

[finnschiermer]

This is a meta-issue used to link up reports of crashes with encryption.

All issues reported after introduction of async writes/commits:

Crash during commit: #5332, realm/realm-swift#7659

Out of bounds during read_barrier: #5230, #5190 (last stack)

Reading table (during read of schema): realm/realm-swift#7640

Decryption failure when reading transaction history: realm/realm-js#4428, realm/realm-java#7652

Initial opening of file (attach): realm/realm-js#4358

Reported before introduction of async writes/commits:

None.

[finnschiermer]

Note: turns out the fix in #5360 is likely not addressing the real issue.

[finnschiermer]

New fix in progress here: #5375

Update: abandoned, real fix below

[finnschiermer]

Fixed by #5383

[Frans-L]

Awesome to see that there's some progress here.

By the way, is this also fixing the underhood issue #5174 ?

[bmt2018]

Great to hear that this issue has been fixed! Any idea when new SDK builds that incorporate this change in core will be released?

[Frans-L]

Hey! Do you have some updates / educated guesses on whether this has been really fixed or not? This issue is still open and most of the linked issues also.

I know that no one has been able to reproduce this bug reliably. So, I understand that it's difficult hard to know whether this has been fixed or not.

However, do you have some estimation of whether this is really fixed, or was the "fix" just an attempt to fix the issue? :)

[sync-by-unito]

➤ Finn Andersen commented:

We consider this fixed by #5383

We have not fixed #5174 which we now believe is theoretical, not of concern on any real device for the single process case. The problem described there for supporting multiprocess with encryption remains and will be fixed as part of building that support.