- Add a package migration manager to be able to migrate external
packages into N1, and decide if they should be enabled or disabled by
default when added.
Summary:
It seems that #2518 was being caused because `fs.writeFileSync` ocasionally failed and ended up writing an empty config.json file, effectively blowing away your N1 settings. The most common cause of failure for `writeFileSync` seems to be EBUSY error on win32, according to Sentry. The reason as to why EBUSY is happening so frequently on win32 is still unclear.
This commit:
- Adds an `atomicWriteFileSync` helper which writes to a temporary file first, and then renames the file. If writing to the temp file fails, we wont blow away our actual config.json. Also, renaming seems to be atomic across all platforms.
- Adds more robust error handling and messaging to both loading and writing the file, and report them to Sentry
- Automatically retries saving the file 3 times before telling the user.
- Fixes https://sentry.nylas.com/sentry/edgehill/group/43112/
Test Plan: Manual
Reviewers: evan, bengotow
Reviewed By: bengotow
Differential Revision: https://phab.nylas.com/D3245
Summary: Future services that require OAuth get a cute new component that lets them connect more easily.
Test Plan: Tested manually.
Reviewers: juan, evan
Reviewed By: evan
Differential Revision: https://phab.nylas.com/D3186
Summary:
- There are some cases in which constantly retrying send can cause unexpected bugs like sending multiple times, so don't retry send at all
- Make 429 a permanent error code
Test Plan: Manual
Reviewers: jackie, evan
Reviewed By: jackie, evan
Differential Revision: https://phab.nylas.com/D3177
Summary:
Adds a new notification bar for mac users only that warns you if N1 is not
in your /Applications folder. Gives you the option to move it to your
/Applications folder.
Test Plan: manual
Reviewers: juan, jackie
Reviewed By: juan, jackie
Differential Revision: https://phab.nylas.com/D3166
Summary:
See edgehill revision: https://phab.nylas.com/D3160
- CustomContenteditableComponents.get now takes a single component to register as an overlaid component, and this component will be passed a prop `{isPreview: true}` when it is meant to be rendered as a preview. Update scheduler and mail-merge to use this new api.
- This prevents the mail merge state from being incorrectly updated when we toggled preview mode -- previously, it believed that the tokens had been unmounted when toggling preview, so they were incorrectly removed from the mail merge state.
- Additionally, makes sure that tokens are correctly added to the mail merge state whenever a token is rendered via componentDidMount -- previously, we only added the token to state when a token was dropped on the contenteditable, however, a token might be pasted or an anchor node might be added by any other means, in which case we want the mail merge state to reflect that. (Adds docs about this)
Test Plan: Manual
Reviewers: jackie, evan
Reviewed By: jackie, evan
Differential Revision: https://phab.nylas.com/D3161
Summary:
fix(subject-field): Fix subject text field focus
Before, if you clicked a mailto link, once the composer loaded, the subject field (which was focused in composer-preload) blurred. This was not ideal as users would want to type once the popout loaded and instead they would have to click first. We identified that this was coming from the composer header _renderSubject injected component. We fixed the focus within mail merge subject text field and additionally added the onComponentDidChange method to injected component. This allowed us to only update the component when the header field changed.
Test Plan: tested on my machine for mailto links
Reviewers: juan
Reviewed By: juan
Subscribers: juan
Differential Revision: https://phab.nylas.com/D3150
Summary:
Fix to pass N1 ID auth through to edgehill-server for remaining services
(clearbit and keybase) that request authing against the Nylas API. We need
to pass along the N1 ID.
This also requires an Edgehill Server patch here:
https://phab.nylas.com/D3149
Test Plan: Manual.
Reviewers: bengotow, juan
Differential Revision: https://phab.nylas.com/D3151
Summary: Ran into issues with the selected tag having different styles associated with it in taiga and in darkside. Fixed these.
Test Plan: Tested on my machine
Reviewers: jackie
Reviewed By: jackie
Subscribers: jackie
Differential Revision: https://phab.nylas.com/D3144
* Added support for 24-hour time to the thread list view (Issue #682)
* Add 24-hour time support to the thread list scroll tooltip (Issue #682)
* Fix for 24-hour time in the thread list scroll tooltip (#682)
Correctly imports the DateUtils module
* Add support for 24-hour time to the draft threads list (Issue #682)
* Add 24-hour time to the message sidebar
* Fix for 24-hour time in the message view so the rollover tooltip is 24hour also (#682)
* Removed unused date functions from utils
fullTimeString and shortTimeString from src/flux/models/utils.coffee were not
compatible with 24-hour time. These functions were modified and moved to
DateUtils in src/date-utils.
* Fix for display of 24-hour time in the message view (Issue #682)
* Removed unused import of Utils in a couple of files
Prompted by Travis build errors.
* Updates to handling of date/time display
Incorporates changes suggested by @bengotow.
Re-enables support for the isDetailed property in message-timestamp (if this is set
to true, a medium length date/time string will be used for display).
Re-enables additional display varieties based on when the email was received. Note
that this is implemented slightly different to the orinal version - time is now given
as an absolute time rather than "... days ago" format.
TZ guessing moved to the global scope of date-utils for performance reasons.
* Minor de-linting
* Re-enable all tests by unfocusing the test suite
A previous commit (ad04775) added an fdescribe() to one of the tests in
draft-helpers-spec. This changes that to a regular describe() so that
all tests will be run when running ./N1 --test.
* Added tests for the new DateUtils functions
Added tests for getTimeFormat, mediumTimeString and fullTimeString.
Removed no longer relevant tests from message-timestamp-spec as _formattedDate
has been removed in favour of the functions in date-utils.
To test shortTimeString, we need to be able to set a fake current time which is
possible in jasmine 2.0+ but not in 1.3 which is currently in use.
As a possible bug, when running more than 10 tests the following warning is raised:
"(node:25025) Warning: Possible EventEmitter memory leak detected.
11 on-config-reloaded listeners added. Use emitter.setMaxListeners() to increase limit",
source: internal/process/warning.js (24)
* Minor de-linting
Summary: After importing an uneven csv (not all rows have the same number of columns e.g. rows: [[1,2][3,4,5]]), deleting the last column would improperly delete data in rows with fewer columns. This is because we were slicing each row based on row.length - 1 which is not the same for all rows. Fixed this by creating a new variable newNumColumns which is columns.length - 1.
Test Plan: Added new test data with uneven data and added a test in table-data-source-spec for removeLastColumn on this uneven set.
Reviewers: juan
Reviewed By: juan
Subscribers: juan
Differential Revision: https://phab.nylas.com/D3133
