Fixes the following issues
- https://chatwoot-p3.sentry.io/issues/5966466083
- https://chatwoot-p3.sentry.io/issues/5966497518
- https://chatwoot-p3.sentry.io/issues/5966555379
For the first one, I am not sure if the fix is 100% correct, since I was
not able to reproduce, but I am confident it will work.
For both, 1st and 2nd issues, the problem came from the fact that we set
individual records to `undefined` when the intent was to remove it,
explicitly using delete fixes the issue.
### Whats up with the store changes?
Glad you asked, this comes down to Vue reactivity, previously Vue didn't
track undefined strictly, it just kinda ignored it, in Vue 3, the
reactivity changed significantly when they introduced ES6 proxies. The
Proxy tracks all property changes, including those set to undefined, so
properties remain enumerable.
So to delete a record, we actually have to the delete it using the
delete keyword, or replace the parent object with a new object splicing
the object to be deleted out.
I am assuming it worked earlier because VueX 3 reactivity was using
Object.defineProperty. Setting it to undefined might have "deleted" it
earlier
---------
Co-authored-by: Pranav <pranav@chatwoot.com>
We will use
[histoire](https://histoire.dev/guide/vue3/getting-started.html) for
component development. I've locked the version to 0.17.15 as it had
issues in the latest versions.
Run the following commands to start the development server.
```bash
# Start the development server
pnpm story:dev
# Build the assets to deploy it to website
pnpm story:build
# View the production version of the build
pnpm story:preview
```
This PR updates the sidebar component to use a native <a> tag for the Help Center URL component. It also updates the build pipeline to use the esbuild options minifyIdentifiers and keepNames set to true.
In the `DropdownList.vue` component, the `onSearch` function was not properly passing the search value to the parent component. This resulted in the `onSearch` event in parent components (such as `LinkIssue.vue`) receiving an undefined value instead of the actual search term.
f18ed01eb7/app/javascript/dashboard/components/ui/Dropdown/DropdownList.vue (L45-L52)
The issue was resolved by modifying the `onSearch` function in `DropdownList.vue` to correctly pass the search value to the `debouncedEmit` function:
When moving form using Gmail Legacy auth to using OAuth, we need the
email address that will be used to connect. This is because we need to
store this email address in the cache and reuse when we get the callback
to find the associated inbox.
However there are cases where the imap login might be
`support@company.com` and the email used to communicate will be
`contact@company.com` (Probably an alias) In that case, we need to send
the correct email address to Chatwoot when re-authenticating
At the moment, we used the inbox email. This PR adds a check that
defaults to to `imap_login` if that is available and imap is enabled
This PR also fixes an unrelated problem where the email inbox creation
flow was not working
---
Tested it, it is working correctly
On production on multiple instances it may happen that the UI is
rendered in correctly, with a lot of options in the sidebar not
available. On further investigation I found out that the feature flag
checks were disabling multiple of those, and also we could see many
correlated errors that pointed towards missing information.
So, there were two problems here
1. The `vuex-router-sync` was not very reliable in some cases
2. In `App.vue` the watch on `currentAccountId` didn't always trigger.
## Fix Tested on Staging
Basically tried to reload the page ~50 times with cache enabled,
disabled, throttling, navigating different pages.
https://www.loom.com/share/1bb27294aa364ac4acfb647780d6385a?sid=87e31330-8cb7-4ded-8616-5e95e2ae3516
<details><summary>
#### What I thought was the fix
</summary>
<p>
### My chain of actions
Replacing vuex-router-sync at first worked fine, but then I saw it was
still failing in some cases, I assumed (I was half-correct tho) that the
rendering of the `App.vue` and syncing of the route to the store was not
happening in a synchronous pattern. So I decided, let's not rely on the
store when the route is directly available in the App context.
Following this, I refactored `useAccount` composable to use `useRoute`
directly, instead of the store, and then replaced the getter inside
`App.vue`. What this did was surface the issue but more consistently 🤯
I saw the watcher, added some console logs, and turns out it was not
getting triggered in all those cases. So I added an `immediate` to it.
And viola, it works!
At the moment, this is deployed to staging and seems to be working
correctly. But we still need to verify it for sure, since how this issue
was surfaced is still a mystery. All we know is that it shows up when
the widget is also loaded alongside the app (if it loads before or after
the app, it works fine)
### What about the route in the store?
Well I have used the `route` usage there with fallback to the store
state. Since Vuex exists in the app context, the route should always be
available to it. But after today I have lost all trust in JavaScript and
will worship rails until end of my life, so I added that in a
`try-catch` block, logged the error to Sentry
</p>
</details>
## Here's the real fix
If you read the explanation I wrote earlier, I thought I fixed the
issue, but then the chat list navigation completely broke. So I removed
the custom route sync implementation and added the original package
back. Turns out the vuex-router-sync earlier was placed after the app
was initalized, however for it to work, the vue app context is not
required. And it's best to run it before the app is even bootstrapped,
so I added it back and placed it correctly.
So the following changes fixes this problem
1. Hoisting the `sync` function call to before we call `createApp` this
ensures that the stores and route hooks are in place before even the app
is created
2. Ensuring the `initializeAccount` is run immediately when watching
`currentAccountId`
4. Source `currentAccountId` for critical top of the tree components
directly from the route instead of the store
There were two warnings showing up on new teams page
1. `errorMessage` prop was getting invalid value, this was because were
short circuting the error message using `&&`, fixed it by using ternary
operator
2. `vue-i18n` has deprecated [rails
syntax](https://vue-i18n.intlify.dev/guide/essentials/syntax#rails-i18n-format),
fixed that warning by removing `%` before `{}` for interpolation
> Note: the `vue-i18n` deprecation needs to be handled, but we can do it
later
This PR has the following fixes
1. preview of article inside the iframe, earlier this didn't work because the iframe didn't get the URL correctly. We fix that by passing the URL is a query instead
2. Emoji picker caused a weird redirect, this was only happening when the chunk was loaded async, this PR changes it to use regular loading instead
When Vue 3 is used with options API, any assignment to `this.<something>` is converted to a Proxy before assignment. This is fine as long as we are in the options context, problem arises when we access this in a `composable` any mutations on the object doesn't behave correctly as expected, this PR fixes that by moving the `automation` object inside the composable and using it in the options.
> Another option to fix such an issue is to make the object non-reactive, like done in places where we have `editorView`, but that wasn't viable here
---------
Co-authored-by: Muhsin Keloth <muhsinkeramam@gmail.com>
This PR makes the following changes
1. Update v-model bindings for components using the old `value` prop and `input` event method
2. Remove components that were not used anywhere
---------
Co-authored-by: Muhsin Keloth <muhsinkeramam@gmail.com>
Vue i18n would automatically merge arrays previously, it does not do so now. This PR fixes it by cleaning up the payload before passing it for translation
The `initOnEvents` was used to get the notification sound file and
trigger the 30 second loop, but since the function was replaced to using
class syntax, the removeEvent listener was not working. This PR fixes it
by reverting to the old syntax but moving it inside the constructor
instead and also adding a `once: true` to ensure it is always removed
automatically
These fixes are all auto generated and can be merged directly
Fixes the following issues
1. Event used on components should be hypenated
2. Attribute orders in components
3. Use `unmounted` instead of `destroyed`
4. Add explicit `emits` declarations for components, autofixed [using
this
script](https://gist.github.com/scmmishra/6f549109b96400006bb69bbde392eddf)
We ignore the top level v-if for now, we will fix it later
This PR allows migration of legacy GMail inbox users to new OAuth based
inbox
## How to test?
1. Create an inbox from the seed data and set it's IMAP address to
`imap.gmail.com` from the UI
2. Open `rails console` and run the following
```
inbox = Inbox.find(100) # use your inbox id here
channel = inbox.channel
channel.update(email: 'hello@chatwoot.com')
channel.prompt_reauthorization!
```
3. This will show the prompt on the UI. Once you click on Reauthorize,
it should open Google Auth. Reauthroize with the same email address as
used in the inbox and it should start working as usual
### Setting up ENV
```sh
GOOGLE_OAUTH_CLIENT_ID=<some-hash>.apps.googleusercontent.com
GOOGLE_OAUTH_CLIENT_SECRET=<client-secret>
GOOGLE_OAUTH_CALLBACK_URL="http://localhost:3000/omniauth/google_oauth2/callback"
```
This PR fixes the issue where proper error messages from the backend
were not displayed when an email already exists in the system during a
profile update, or when a phone number is already taken for Twilio
during the creation of a new account.
Fixes:
https://linear.app/chatwoot/issue/CW-3560/prod-customer-facing-issue-updating-email
Fixes https://github.com/chatwoot/chatwoot/issues/6228
There is a country code selector for the phone input field. This is
often a point of frustration. See the response below.
> We are using the phone number field however this can be frustrating
for customers, especially mobile users, to select the +1 US country
code. Our users are typically local businesses and being able to default
to +1 country code in the phone number field or the account would
improve the interaction they have with customers.
Most people who run local businesses don't need a country selector.
However, to preserve the quality of the data we store, we need a country
code.
A balance between these two issues can be found with an auto-populating
country code field based on the browser's timezone. This is what I did
in this PR. Based on the browser timezone, we will resolve it to the
closest country code.
This PR enhances the emoji search functionality in the editor's emoji selector by improving how emoji names are displayed and searched.
The UI now shows emoji names instead of slugs, and the search logic has been updated to generate `searchString` without whitespaces, allowing users to search for emojis like 'face_with' using 'facewith'
