This PR moves the size comparison action back to a separate workflow
which now uses the `pull_request_target` event. This event is triggered
at all the same times as the `pull_request` action, except that the
workflow file content comes from the target branch of the PR, and it is
run in the context of the repo owning the target branch. Practically,
this means that it will still have access to post a comment even if the
PR comes from a fork.
We don’t want the build actions to be run in a `pull_request_target`
workflow because they would get access to the secrets and be able to
perform arbitrary actions on the repository, even from fork PRs.
See the current version failing here:
https://github.com/actualbudget/actual/actions/runs/5395184895/jobs/9797388016?pr=1122
This significantly speeds up `yarn lint` for me. It also ensures we’re
listing all source files in the project, including the `.eslintrc` files
and any other files that may be present.
It me it seems the monkey patch is no longer necessary.. At least I'm
not able to reproduce any `escape` key related issues. But LMK what you
think.
The `@reach/listbox` component is used for the custom select component
(for example: for filters).
I've updated the return values in `sync/index` to sensible values based
on how they're being called.
I've updated the type of `msg.value` to be `string | number | null` to
match the values expected by `serializeValue`
- New linter rules are now added
- Perhaps the `createPayee` method i've tagged with a disable should be
deleted.
- I have ignored unused methods in Plaid/Mobile as this stuff is still
WIP/experimental?
None of this code is relevant to people who have used the open source
app. I initially wanted to leave it in so it could be used for major new
features in the future, but as it falls further and further out of date
I think it’s best to just delete it in a clean PR that can be referenced
later if desired.
Closes#1069
I've not actually tested this change. Which is why I'm not changing it
to direct links (as they might not work).
Instead I'm just applying a very quick patch so the message would not be
misleading anymore.
This PR is for the first item listed in #559: `Expand All / Collapse All
Categories`
For the expand / collapse all categories functionality, I was choosing
between having a single `Expand / collapse all categories` button or one
for each: `Expand all categories` and `Collapse all categories` buttons.
For the initial implementation, I have opted with the latter. Please let
me know which one is the right way to go or if there are other
suggestions and I'll just accordingly.
actual-server does not need to import the full actual-app/api package.
It can import only the CRDT stuff.. so I'm extracting it into a new
package to reduce the size of actual-server and make the link between
things more transparent.
Continuing on from #1139 this PR deletes all function that are not used
anywhere.
The next PR will include all the entire files that are unused & deleted.
This reverts #1029. As raised in #1097, the formatting chosen doesn’t
work well when doing math. There may be a way to balance compatibility
with multiple format styles with handling non-currency amounts
correctly, but it will require some more careful consideration. Re-opens
#894.
URLs with the trailing slash don’t work well — requests end up being
made to `https://example.com//sync/sync` and such which can 404
---------
Co-authored-by: Matiss Janis Aboltins <matiss@mja.lv>
As discussed in #1126 creating separate PRs.
The linter rule isn't actually part of this first PR and will likely be
in the final PR as otherwise it will just fail because all the issues
aren't resolved.
The code is 'ready' though so as soon as this is merged I will open next
one(s)
Closes#1125
Disables (hides) the "all payees" checkbox when schedules are loading.
Reproduction:
1. open `/schedule/discover`
2. while the data is still loading - click on the checkbox in the table
header
3. after loading finishes - the page crashes with an error
Previously, the frontend would attempt to make real requests to
`https://not-configured/`, which of course failed. I’ve changed the
internal structure to have the lack of a server expressed as a `null`
server. A new `did-bootstrap` browser-level setting has been added to
track if the user clicked “Don’t use a server.”
Finally, I updated the auth logic for methods that call out to the
server to make them consistent and use the early return style which
reduces indentation.
