The CSS is now served as an external file instead of being injected.
This was done to improve performance. Ideally we would come to a middle
ground between this and the former behaviour by injecting only the CSS
that was required for the initial page load, then lazily loading the
rest. However that change would be more complex. The hope was that
making all CSS external would at least be a slight improvement.
Performance metrics were collected before and after this change to
determine whether this change actually helped. The metrics collected
were the timing events provided by Chrome DevTools:
* DOM Content Loaded (DCL) [1]
* Load (L) [2]
* First Paint (FP) [3]
* First Contentful Paint (FCP) [3]
* First Meaningful Paint (FMP) [3]
Here are the results (units in milliseconds):
Injected CSS:
| Run | DCL | L | FP | FCP | FMP |
| :--- | ---: | ---: | ---: | ---: | ---: |
| 1 | 1569.45 | 1570.97 | 1700.36 | 1700.36 | 1700.36 |
| 2 | 1517.37 | 1518.84 | 1630.98 | 1630.98 | 1630.98 |
| 3 | 1603.71 | 1605.31 | 1712.56 | 1712.56 | 1712.56 |
| 4 | 1522.15 | 1523.72 | 1629.3 | 1629.3 | 1629.3 |
| **Min** | 1517.37 | 1518.84 | 1629.3 | 1629.3 | 1629.3 |
| **Max** | 1603.71 | 1605.31 | 1712.56 | 1712.56 | 1712.56 |
| **Mean** | 1553.17 | 1554.71 | 1668.3 | 1668.3 | 1668.3 |
| **Std. dev.** | 33.41 | 33.43 | 38.16 | 38.16 | 38.16 |
External CSS:
| Run | DCL | L | FP | FCP | FMP |
| :--- | ---: | ---: | ---: | ---: | ---: |
| 1 | 1595.4 | 1598.91 | 284.97 | 1712.86 | 1712.86 |
| 2 | 1537.55 | 1538.99 | 199.38 | 1633.5 | 1633.5 |
| 3 | 1571.28 | 1572.74 | 268.65 | 1677.03 | 1677.03 |
| 4 | 1510.98 | 1512.33 | 206.72 | 1607.03 | 1607.03 |
| **Min** | 1510.98 | 1512.33 | 199.38 | 1607.03 | 1607.03 |
| **Max** | 1595.4 | 1598.91 | 284.97 | 1712.86 | 1712.86 |
| **Mean** | 1553.8025 | 1555.7425 | 239.93 | 1657.605 | 1657.605 |
| **Std. dev.** | 29.5375 | 30.0825 | 36.88 | 37.34 | 37.34 |
Unfortunately, using an external CSS file made no discernible improvement
to the overall page load time. DCM and L were practically identical, and
FCP and FMP were marginally better (well within error margins).
However, the first paint time was _dramatically_ improved. This change
seems worthwhile for the first paint time improvement alone. It also
allows us to delete some code and remove a dependency.
The old `css.js` module included two third-party CSS files as well, so
those have been imported into the main Sass file. This was easier than
bundling them in the gulpfile.
The resulting CSS bundle needs to be served from the root because we're
using a few `@include` rules that make this assumption. We could move
this under `/css/` if desired, but we'd need to update each of these
`@include` rules.
Relates to #6646
[1]: https://developer.mozilla.org/en-US/docs/Web/Events/DOMContentLoaded
[2]: https://developer.mozilla.org/en-US/docs/Web/Events/load
[3]: https://developers.google.com/web/fundamentals/performance/user-centric-performance-metrics
The Babel config had previously supported all browsers with greater than
0.25% global usage (according to `browserlist`). This resulted in
`babel-preset-env` including plugins sufficient to support the following
minimum browser versions:
```
{
"chrome": "49",
"android": "4.4",
"edge": "16",
"firefox": "52",
"ios": "9.3",
"safari": "11"
}
```
Instead, the babel config now explicitly supports chrome >= 58 and
firefox >= 53. Chrome and Firefox are the only browsers we currently
publish to, and these were the minimum versions with no additional Babel
transformations.
The minimum browser versions we support should be re-evaluated later,
when we have added tests and documentation.
The plugin 'transform-async-to-generator' has also been removed. It was
used to translate async/await, but our browser targets all support
async/await.
Removing some of these transformations exposed bugs in `uglify-es` that
only presented themselves in the production build. `gulp-uglify-es` has
been updated to a version that uses `terser` instead of `uglify-es`,
which has resolved these issues.
Relates to #6805
Set the minimum browser version supported in the extension manifest.
Currently we only ship the extension on Chrome and Firefox, so the
minimum version has been set for those two browsers.
Relates to #6805
* ci: Rename full_test to test_and_release
* ci: Add scripts to automate GH releases
* Add .bak files to .gitignore
* ci: Add reviewer to the auto version PR
* Abstract domain provider from its stream transport
Creating new provider-consuming extensions, like [a new
platform](https://github.com/MetaMask/metamask-extension/blob/develop/docs/porting_to_new_environment.md)
can be frustrating for new contributors because our provider
construction has been tangled with our streaming interface.
Here I've broken up our streaming domain connection from the provider
construction, so developers can more easily construct local and
domain-restricted providers without dealing with streams.
* Abstract public API from stream interface
* clean up noop
* Document non-streaming interface
* getSiteMetadata must be async
* Clean up filters on stream end
* Document cleaning up filters
* Allow named filterMiddleware to be cleaned up
* Linted
* Require site metadata
* Destroy any destroyable middleware during cleanup
* Lint
The Drizzle tests have not been used for some time. They were used to
ensure compatibility with newer versions of `web3` v1. If we want to
re-add tests to ensure compatibility with newer `web3` versions, we
should find some way of doing that more reliably than was done here -
these tests were somewhat flaky and unreliable.
Update lodash transitive dependencies in response to security advisory:
https://www.npmjs.com/advisories/1065
There are a few remaining instances of the vulnerable lodash dependency
in the lockfile, but those are only used by development dependencies.
They are pinned, so are more difficult to update.
These files were referencing npm scripts that no longer existed. Notices
appear to no longer exist, and the `ui-dev.js` module is no longer
actively used.
The `mock-dev.js` module is still used for certain integration tests, so
I've just removed the reference to the non-existent script.
We've been using the `eslint-plugin-json` plugin for some time, but we
haven't been visiting `.json` files in the lint script. The lint script
has now been updated to incude `.json` files, which means any invalid
JSON will result in a lint error.
Unfortunately this JSON plugin doesn't seem to apply the other eslint
rules (such as `key-spacing`) to the JSON files. I wasn't able to find a
way to get that to work. Instead I manually auto-formatted each of the
locale `message.json` files, which fixed many whitespace
inconsistencies.
The `states.json` file was deleted completely, as it appears to be
unused. It wasn't a valid JSON file anyway, it was JavaScript. It looks
like a `states.js` file is automatically generated, but an old copy was
accidentally saved as `states.json` and included in the repo.
Many duplicate key errors were found and fixed in the
`development/states/` JSON files.
`package-lock.json` was added to `.eslintignore` because it was very
slow to lint, and linting it doesn't provide much value.
We had forgotten to add `eslint` as a dependency, even though we use it
directly. It had always worked because we have dependencies that also
depend upon it.
`eslint` has also been updated to v6, which necessitated two minor
changes.
The AppVeyor configuration appears to be unused - this project is not
connected to AppVeyor.
After deleting the AppVeyor config, the JSDoc config was the last thing
in the `development/tools` directory. That felt a little silly, so I
moved it up to `development`.
Unused expressions are generally a mistake, as they don't do anything.
The exceptions to this rule (short-circuit expressions and ternary
expressions) have been allowed.
The `webrtc-adapter` was previously ignored by eslint because it has a
side-effect upon being imported. I removed the local variable instead,
which should preserve the same side-effect without making eslint
complain.
* Add React and Redux DevTools
* Conditionally load react-devtools
* Add start:dev npm script to run the app with devtools
Co-Authored-By: Mark Stacey <markjstacey@gmail.com>
* Improve auto changelog script
The auto changelog script was creating empty or invalid entries in a
number of different cases, such as when the body of a commit spanned
multiple lines. This has been fixed, and the following additional
improvements have been made:
- Error handling (it will now crash upon encountering an error)
- Commits without a PR number in the subject are listed without the PR
prefix
- Invalid shellcheck warnings ignored
- Only the first line of the commit body is shown
- Carriage returns are stripped (some commits contain them)
This script should be more reliable for helping to manually update the
changelog. It's still not sufficiently robust to use as part of an
automated process - I don't think that's feasible without maintaining
stricter control over commit messages conventions and/or merge
strategies.
* Add loading spinner to pending tx status label.
* Add border around account icon in top right
* Change style of settings toggle buttons; wrap with local components
* Eliminate large space after settings labels when no description
* Remove network form from advanced tab of settings
* Keep new account container height to contents when in full screen
This package is deprecated, and is incompatible with React v16. It has
been replaced by `react-test-renderer`, which has a drop-in replacement
for `react-addons-css-transition-group`.
Strangely, `react-test-renderer` was already listed as a dependency
despite not being used. I had to downgrade it, as the version already
listed was for React v16, and we're still using React v15.
This must have been removed as a dependency at some point, but left in
the lockfile. I noticed it would install itself whenever I reinstalled
all packages.
The `stage-0` Babel preset has been replaced with the specific Babel
plugins that we depend upon. We don't use most of `stage-0`, so this
allowed us to remove many unnecessary transformations. We had to remove
this preset soon anyway, because all of the stage presets are
deprecated in Babel 7.
The `stage-0` preset consisted of these plugins:
```
"transform-do-expressions"
"transform-function-bind"
"transform-class-constructor-call"
"transform-export-extensions"
"transform-class-properties"
"transform-decorators"
"syntax-dynamic-import"
"syntax-trailing-function-commas"
"transform-async-generator-functions"
"transform-async-to-generator"
"transform-exponentiation-operator"
"transform-object-rest-spread"
```
Of that list, only 'transform-class-properties', 'transform-object-
rest-spread', and 'transform-async-to-generator' were being used.
While working on #6805, I noticed that many variables were being used
before they were declared. Technically this worked fine in practice
because we were using the `transform-es2015-block-scoping` Babel plugin,
which transforms `let` and `const` to `var`, which is hoisted. However,
after removing that Babel transformation, many things broke.
All instances of variables or classes being used before declared have
been fixed.
The `no-use-before-define` eslint rule has been added to catch these
cases going forward. The rule is disabled for function declarations for
the moment, because those are always hoisted. We could disable that too
if we want to, but it's purely stylistic and would require a lot more
changes.
* Added visual documentation of the menu bar component
* updated CHANGELOG.md
* renamed to account menu which is the component highlighted in the screenshot
* updated contributions document
* Remove unused state 'gas.basicPriceAndTimeEstimates'
* Check for invalid estimates from local storage
Gas estimates were being cached in local storage then later retrieved,
but the retrieved values were not being checked. If the data failed to
save, failed to load, or was cleared since being saved, it would result
in the gas estimates being set to undefined.
The estimates retrieved from local storage are now checked before they
are used. If they are falsy, the estimates are retrieved from the
network instead.
This should fix this Sentry issue:
[METAMASK-6W0T](https://sentry.io/share/issue/cfe470314a5741768b19050815322aa4/)
A few additional changes were made to the gas-duck tests to accommodate
the use of `sinon.restore`. `restore` is strongly recommended by the
`sinon` team, as neglecting to use it can result in memory leaks. It has
the additional benefit of ensuring you create fresh stubs/spies for each
test, which means they no longer need to be reset between tests.
* Introduce delay for eth_estimateGas calls with in test
* Add test that fails when gas estimates of contract method calls without gas are too high.
* Get transaction gas data from unApprovedTxs instead of confirmTransaction
* Fix selection of gas data in gas-modal-page-container.container
* Lint changes related to Version-6.7.2-gasLimitFix
* Fix e2e tests on Version-6.7.2-gasLimitFix
* Fix unit and integration tests for changes from Version-6.7.2-gasLimitFix
* more e2e fixes
* Add assertions for transaction values on confirm screen
* Fix display of transaction amount on confirm screen.
The `disc` gulp command no longer works. I wasn't able to fix this
easily, so instead it has been removed. We can probably find something
better to replace it with.
`addons-linter` will occasionally run out of heap space. This provides 3
GB of heap for that script rather than the default ~1.5 GB. The CircleCI
containers have 4GB of memory, so this should leave plenty of extra
space for non-heap memory.
Alessandro Ricottone
Whymarrh Whitby
Mark Stacey
Dan Finlay
Terry Smith
Thomas Huang
Erik Marks
Saxon Knight
filmendless
Brian Soule
ryanml
Dan J Miller
Mark O'Sullivan
Dan Finlay