Re-add # shellcheck source=/dev/null

Synced with @jon offline. We came to the conclusion that bash does not close the double quote in "$( until it encounters a )", so we're safe to use a double quote inside the subshell

I still do not understand if it's dangerous to leave the variable unquoted. You and Shellcheck both seem to think that bash variables need to be quoted, but you are both suggesting a solution where it is not quoted. Can you please clarify why you think it's safe to leave the variable unquoted?

You addressed a lot of things, but you didn't address the thing I asked for re-review on. Can you respond to this please?

I wish you had deferred this work instead of randomly picking it up. You probably thought it would be easy, but now you have a diff and going to have to spend even more time on this responding to comments...

I agree we should only be changing calendar filters in one place. Can we move the logic for changing them here to lib?

@jon – ping on this, please make sure to monitor your review queue!

cc @anunay – @marcin already has a diff up with X3DH, one day after we landed it! Exciting

@atul has recently worked on the permission migration stuff and is more familiar with the SQLite redux-persist migration stuff, so will defer to him

This is safe because uploads has an index on container (would be good to note this in your Test Plan)

This is safe because uploads has an index on container (would be good to note this in your Test Plan)

Please do not land if this could potentially lead to us publishing a build that has a button that does nothing

Where is this used? Can you include the usage in this diff please? It's hard to review these standalone functions without seeing how they're used

Where is this used? Can you include the usage in this diff please? It's hard to review these standalone functions without seeing how they're used

Please make sure CI passes before landing

Use toBeDefined(), rebase

Update to 0.0.4

Update to 0.0.4

Rebase

Update to 0.0.3

Update to 0.0.3

Add unpublished_prekey

Great catch on the descendant permissions! Glad we did this testing

Add declare

Extracting to a bash file makes sense... it's way more readable, and we're able to get Shellcheck to run on it. That said, Shellcheck was a bit confused on one line (I think) – would love to get @jon's take

Extract to a bash file

@rohan, @atul just finished testing D7364 and might have some pointers... it looks like Redux dev tools don't work particularly well

Nice, let's land it!! Thanks for doing that

Perfect!! As a last step before landing, can you try the test I described to @atul here? To run the Redux dev tools that I reference there, check out this doc. And the way you can see the state check mechanism being run is you should see a PROCESS_SERVER_REQUESTS action. If there's just one, that means the state check mechanism ran and saw no changes. If there are multiple such actions in a row, that means the state check mechanism detected changes. You should wait around 3-5ms after the socket connects to see the state check mechanism fire, but you might need to manipulate sessionCheckFrequency to get it to fire more frequently than once every 3 min (if you need to run it again soon after running it once).

In D7364#219120, @atul wrote:

Eg. compile prod build that doesn't support avatar permission with old codeVersion, point it at local keyserver, run keyserver that optionally omits avatar permission – make sure state check mechanism is being called via redux-devtools (hopefully it works better for you than me), and that it isn't detecting any changes

Not sure I completely follow?

Are you suggesting that I:

1. Locally update role-creator.js to exclude avatars
2. Run updateRolesAndPermissionsForAllThreads migration
3. Compile native build and connect to keyserver so I'm missing all thread_edit_avatar permissions
4. Add avatar permissions back to role-creator.js
5. Run updateRolesAndPermissionsForAllThreads migration
6. Observe whether there are any state check discrepancies when native re-connects to keyserver?
7. Manually set db_version of metadata table back to whatever it is on master so I'm "back in sync" with migrations

Last round of changes!

JS seems to follow from Rust changes. Docker changes make sense

Defer to @jon, most of this is Rust and the rest of it follows from the Rust

Overall looks good, but we need to get better at maintaining a consistent design system

It would be good to amend the Test Plan to include checking the state check mechanism. Eg. compile prod build that doesn't support avatar permission with old codeVersion, point it at local keyserver, run keyserver that optionally omits avatar permission – make sure state check mechanism is being called via redux-devtools (hopefully it works better for you than me), and that it isn't detecting any changes

Accepting but I think my comment needs to be addressed before landing