Previously, all errors encountered during testing or production were
sent to the primary `metamask` Sentry project, whereas development
errors were sent to `test-metamask` instead. This change ensures that
errors encountered during tests are sent to `test-metamask` as well.
In the case where the initial spend limit for the `approve` function
was set to the maximum amount, editing this value would result in the
new limit being silently ignored. The transaction would be submitted
with the original max spend limit.
This occurred because function to generate the new custom data would
look for the expected spend limit in the existing data, then bail if
it was not found (and in these cases, it was never found).
The reason the value was not found is that it was erroneously being
converted to a `Number`. A JavaScript `Number` is not precise enough to
represent larger spend limits, so it would give the wrong hex value
(after rounding had taken place in the conversion to a floating-point
number).
The data string is now updated without relying upon the original token
value; the new value is inserted after the `spender` argument instead,
as the remainder of the `data` string is guaranteed to be the original
limit. Additionally, the conversion to a `Number` is now omitted so
that the custom spend limit is encoded correctly.
Fixes#7915
* Use ref instead of findDOMNode in jazzicon component
The jazzicon component was using `findDOMNode` to get the DOM node for
the main div returned by the component, which is generally not
recommended. Instead a ref is now used.
* Update Jazzicon to v2
This version drops the dependency upon `raphael`, and no longer uses
the function `createSVGMatrix` which was causing unit tests to fail
(because it's not supported by jsdom).
Our JSDoc documentation has not been updated in a very long time, and
we don't use JSDoc in enough places for the docs to have been
especially useful. The tools and scripts used to generate and publish
these docs have been removed.
References to this documentation have also been removed from the
README.
Hopefully once the TypeScript migration has made substantial progress,
we can generate more useful documentation using something like TypeDoc.
After updating the custom spend limit on the approve screen, the data
for the transaction was not being updated. Instead it showed the
original transaction data. The transaction data was being updated
correctly in the final transaction though.
The approve screen has been updated to ensure changes to the custom
spend limit are reflected correctly in the data shown.
Update accounts permission history on accountsChanged
Create PermissionsLogController
Fix permissions activity log pruning
Add selectors, background hooks for better UX
Make selected account the first account returned
Use enums for store keys in log controller
Add last selected address history to PreferencesController
* Update lodash
All versions of the full `lodash` package have been updated to 4.17.15.
The only exception is v4.17.14 which is pinned by `ganache-core`.
* Switch to using `lodash` instead of per-method packages
We have the full lodash package _ten times_ as a production transitive
dependency, so including per-method packages is not saving space (it
might instead result in slightly more space being used).
The Storybook deploy step is currently broken because it's using the
wrong source branch (`master` instead of `develop`), and because the
key that CircleCI is setup with doesn't have write access to the repo.
While I expect we'll get these two problems fixed soon, this ensures
that we at least have passing builds on `develop` in the meantime.
Unhandled rejections are now caught using built-in Node.js APIs instead
of with `bluebird`. `bluebird` was added as a production dependency but
was only used for this purpose. The code responsible for catching
unhandled rejection in the browser was removed, as this test helper is
never run in the browser.
Additionally, unhandled rejections are tracked over the course of all
tests, and result in a non-zero exit code if they remain at the end.
This was done because it is possible for errors to trigger the
`uncaughtRejection` event but then still be handled later on. This is
uncommon, and doesn't seem to happen in our test suite. But if it does
in the future, it'll be logged but won't result in a non-zero exit
code.
This patch update includes a bug fix for ES6 imports. The bug doesn't
affect our use of this module, but it ensures that we can safely
use the `dedupe` and `bind` APIs as ES6 imports if we decide to.
* Minimum changes to get storybook working
Undo path changes
* Add build:storybook scripts to package.json
* Add storybook deployer
* Add storybook:deploy to package.json
* Update circle ci config
* Update yarn.lock
* Remove addon-info
* Update yarn.lock file to reflect removing of addon-info
Co-authored-by: Dan J Miller <danjm.com@gmail.com>
`node-fetch` is a smaller and simpler project than `request-promise`,
and we already have it as a transitive dependency.
`request-promise` was also incorrectly listed as a production
dependency. `node-fetch` has been added as a `devDependency` to replace
it, as it was only used in one CI script.
Any error caught during a React component render or lifecycle method
will now be caught by the top-level error boundary, which shows the
user this new error page. The error page will display a simple error
message, and will show the details of the error in a collapsible
section.
The caught error is also reported to Sentry.
In development the error will be re-thrown to make it easier to see on
the console, but it is not re-thrown in production.
* Remove unnecessary `getEnvironmentType` parameter
The default value of the first parameter is `window.location.href`, so
there is no need to pass it in explicitly.
* Remove junk parameter from `getEnvironmentType` invocation
`getEnvironmentType` doesn't need to be passed any parameter, as the
default value is `window.location.href` which is generally what is
wanted. In this case, the variable `location.href` was always
`undefined` anyway. This particular `location` variable is from React
Router, and does not have an `href` property.
* Fix comment for `getEnvironmentType`
One of the possible return values was referred to by the wrong name.
`withRouter` has been removed from any components that were not using
any of the three props injected by `withRouter`: `history`, `location`,
and `match`.
`compose` is a no-op when called upon a single component, so it has
been removed in all such cases.
* Add benchmark to CI
The page load benchmark for Chrome is now run during CI, and the
results are collected and summarized in the `metamaskbot` comment.
Closes#6881
* Double default number of samples
The number of default samples was changed from 10 to 20. The results
from 10 samples would show statistically significant changes in page
load times between builds, so weren't a sufficiently useful metric.
The only web API that our usage of DetectRTC relied upon was
'enumerateDevices', which is supported and stable among all of our
supported browsers.
Note that the error handling here is a little... non-standard, and the
logic around how Firefox and Brave are handled should be justified, but
I've left the logic as-is for now to keep this PR small.
The WebRTC spec is fairly stable these days, particularly among the
browsers we support. We don't need this shim for anything. I'm guessing
it may have been added primarily with IE in mind.
The filename is seeded by a simple use of Math.random() pulling from an alphanumeric character bank, as opposed to a more cryptographically random solution. This provides a simple layer of difficulty for bad actors to seek out the recovery phrase file.
A margin of error metric has been added, which is calculated from a 95%
confidence interval. This confidence interval is calculated using
Student's t-distribution, which is generally preferred for smaller
sample sizes (< ~30) of populations following a normal distribution.
The script `benchmark.js` will collect page load metrics from the
extension, and print them to a file or the console. A method for
collecting metrics was added to the web driver to help with this.
This script will calculate the min, max, average, and standard
deviation for four metrics: 'firstPaint', 'domContentLoaded', 'load',
and 'domInteractive'. The variation between samples is sometimes high,
with the results varying between samples if only 3 were taken. However,
all tests I've done locally with 5 samples have produced results within
one standard deviation of each other. The default number of samples has
been set to 10, which should be more than enough to produce consistent
results.
The benchmark can be run with the npm script `benchmark:chrome` or
`benchmark:firefox`, e.g. `yarn benchmark:chrome`.
The 'can retype the seed phrase' test would fail sometimes when one of
the words in the seed phrase was a subset of another word (e.g. 'issue'
and 'tissue'). This is because the selector used to find the word
looked for the first element that contained the text, rather than an
exact match.
To simplify the selector and make it more reliable, test ids were added
to each seed phrase word. The selector now uses CSS instead of XPath,
and it only finds exact matches.
A test id was also added to the div containing the shuffled seed words
to select from, so that the chosen seed words wouldn't be selected
in place of the real target when the same word appears twice.
The `confirm-seed-phrase` component extends PureComponent, so it
doesn't need a `shouldComponentUpdate` function. The state is
effectively immutable, as all state is either a primitive type or is
updated with new instances rather than mutation.
Removing this function will silence a warning message printed to the
console during e2e tests (React doesn't want you to set this function
on `PureComponent`s).
Removing this function also exposed an unused piece of state, which has
also been removed.
There have been intermittent test failures at the beginning of various
e2e test runs. Most tests start with waiting for the 'Welcome' button
to be visible and enabled, which means waiting for the loading screen
to go away.
It looks like the reason the test intermittently fails is that
sometimes the loading screen doesn't appear until a few moments _after_
the page loads (or that it vanishes and comes back).
It was rather difficult to track down each possible cause for the
loading screens, so in the meantime a pause has been added at the start
of each run. This should hopefully suffice to ensure the momentary gap
in loading has been passed by the time the first test starts up.
The `withFixtures` helper will instantiate ganache, a web driver, and
a fixture server initialized with the given set of fixtures. It is
meant to facilitating writing small, isolated e2e tests.
The first example test has been added: simple-send. It ensures that the
user can send 1 ETH to another account.
These new e2e tests will run during the normal e2e test run.
Closes#6548
This reverts commit 4b4c00e94f. The
original change was a possible optimization of CI, though it ended up
not having a huge impact. It was thought that it may have broken source
maps, because the test build overwrote the `dist` directory that the
source maps were written to. However this turned out not to be the
case, as the changes to `dist` are never persisted to the workspace.
This is being re-introduced because the test build is needed for an
additional job (the page load benchmark), and sharing the same build
for all three jobs would surely be faster than building it separately
three times.
The driver now has a page navigation function that can navigate to any
of the three primary pages used in the extension. Additional pages and
support of paths can be added later as needed.
As of #7663, an in-memory store was used in place of local storage
during e2e tests, to facilitate the use of state fixtures. However,
this made it difficult to export state during a test run. The
instructions for exporting state to create fixtures assumed that local
storage was being used.
A new global function has been added to the background context to allow
exporting state. This method is available during testing and
development, and it works with either local storage or the in-memory
store. The fixture instructions have been updated to reference this new
function.
The signature request e2e tests were previously using ropsten. This
expectation was even hard-coded into the test contract dapp.
Instead the contract-dapp has been updated to use the current `chainId`
when calling `signTypedData` (falling back to the `networkId` if
`chainId` is not set). The fixture used by `signature-request` has been
updated to use ganache.
