The web3 usage metrics added in #9144 assumed that all web3 properties
were strings. When a `Symbol` property is accessed, our `inpage.js`
script crashes because the `Symbol` cannot be serialized correctly.
A check has been added for non-string property access. The metric event
in these cases is set to the string "typeof ", followed by the type of
the key. (e.g. `typeof symbol` for a `Symbol` property).
Fixes#9234
The usage metrics for the injected web3 instance were being sent upon
each use, which exceeded the limits of our Matomo plan. These metrics
are now only being sent upon the first usage, for each origin and
property.
The usage metrics for the injected web3 instance were being sent upon
each use, which exceeded the limits of our Matomo plan. These metrics
are now only being sent upon the first usage, for each origin and
property.
* Fix require-unicode-regexp issues
See [`require-unicode-regexp`](https://eslint.org/docs/rules/require-unicode-regexp) for more information.
This change enables `require-unicode-regexp` and fixes the issues raised by the rule.
* Remove case-insensitive flag from regexps
The "Background" metrics category was being set in the
`backgroundMetaMetricsEvent` function. This function might not
necessarily include any event at all though, so setting it here seemed
inappropriate. It would also crash if `eventData.eventOpts` was not
set, which is not great since that property is optional.
The background category is now set in the `sendBackgroundMetaMetrics`
function in `metamask-controller`. This method is used solely for event
data, so it would make sense for this category to be always set.
There is no functional difference, since `backgroundMetaMetricsEvent`
is called solely by `sendBackgroundMetaMetrics`.
The `currentPath` parameter passed to our metrics utility had been
passed the full URL rather than just the path, contrary to what the
name would imply. We only used the path portion, so passing the full
URL did lead to complications.
Now just the `pathname` is passed in, rather than the full URL. This
simplifies the metrics logic, and it incidentally fixes two bugs.
The main bug fixed is regarding Firefox metrics. Previously we had
assumed the `currentPath` would start with `chrome-extension://`, which
of course was not true on Firefox. This lead to us incorrectly parsing
the `currentPath`, so path tracking was broken for Firefox events.
This broken parsing is now bypassed entirely, so metrics should now
work the same on Firefox as on Chrome.
The second bug was that we were incorrectly setting the tracking URL
for background events during tests. As a result, we were incorrectly
detecting ourselves as an internal site that had referred the user to
us. But this was not of major concern, since it only affected test
metrics (which get sent to the development Matomo project).
Lastly, this change let us discard the `pathname` parameter used in
the `overrides` parameter of the `metricsEvent` function. Now that
`currentPath` is equivalent to `pathname`, the `pathname` parameter is
redundant.
* Remove `url` parameter from `metricsEvent`
The `url` parameter was used to override the `currentPath`, but it
never worked correctly. It was supposed to be used for setting the
`url` query parameter that was sent to Matomo, but `currentPath` was
always used even if it `url` was set and `currentPath` was empty.
Instead, `currentPath` is now always used. There was never a need to
provide an "override" for `currentPath` when it can be set directly.
The metrics provider does set `currentPath` automatically by default,
but this can be overwritten already by passing a second parameter to
`metricsEvent`.
There were two places this `url` parameter was being used: background
events, and path changes. Background events were submitted with no
`currentPath`, so because of the bug with the `url` parameter, the
metrics utility would crash upon each event. So those were never
actually sent. This commit will fix that crash.
The `currentPath` parameter was supplied as an empty string for the
path change events, so those never crashed. They just had the `url`
query string parameter set incorrectly (to an empty string). It should
now be correctly populated, which should mean we'll be capturing all
path changes now. Previously we were only capturing path changes to
pages that happened to include an event, because of this blank `url`
problem.
* Use `url` query parameter as fallback for generating `pv_id`
The `pv_id` parameter currently isn't generated correctly on Firefox,
as the generation assumes that the current URL starts with
`chrome-extension://`. The `url` query parameter is still unique for
each path, so it's probably good enough for generating an id for each
page.
This is just a temporary fix; it will be removed in a future PR, where
Firefox will be properly supported.
Background events are now sent in the `Background` category, rather
than `backend`. Conventionally we use the term "background" over
"backend", as it's not really a "backend" in the normal sense since
it's a client background process. Also it's capitalized because all of
the other event categories are capitalized as well.
The metrics URL has been updated to use `background` instead of
`backend` as well, for consistency.
Luckily we don't have to worry about our metrics being disjointed due
to this name change, because the background metrics never worked to
begin with! So there will be none under the old name. The metrics will
be made functional in a separate PR.
The Sentry DSN is now expected to be provided via environment variable
for production builds. The build script will fail if it is missing, and
an error will be thrown at runtime if it is missing.
The `SENTRY_DSN` environment variable has been set in CI to the old
value for `SENTRY_PROD_DSN`. We can migrate to a new DSN at some point
in the future.
In a non-production environment, Sentry was configured to send error
reports to a "test" MetaMask project. It will still do this during e2e
tests, but in development Sentry is now disabled completely.
In practice this was never useful in development.
* Fix popup/notification when browser is in fullscreen, primarily on OSX.
The issue was reported internally via Slack. User was running Mac OSX Chrome in fullscreen mode where Chrome is created in a new Desktop workspace.
The issue reproduced on OSX Chrome in fullscreen/maximized view overrides the explicitly set width and height for `windows.create()`. Possibly not overrides, but creates a window based off of the window that it was created from. Found a related [Chromium bug](https://bugs.chromium.org/p/chromium/issues/detail?id=263092&q=window%20create%20width%20os%3DMac&can=2).
The fullscreen `popup.left` pixel will calculate the window position incorrectly since we set and assume the width of the created window. The incorrect `left` position the window and transition the focus Desktop/Workspace incorrectly and make is seem to lose focus of the new window/workspace. Incidentally this will make the popup full width/height, and create a new workspace for the view, which we have no control over until Chrome
fixes it.
This will check if the popup is 'fullscreen', which it gets passed from the origin window, if so then don't reposition the window. If Chrome fixes the issue we can revert this change.
* Feedback commit
Co-authored-by: Mark Stacey <markjstacey@gmail.com>
Co-authored-by: Mark Stacey <markjstacey@gmail.com>
The `currentPath` parameter passed to our metrics utility had been
passed the full URL rather than just the path, contrary to what the
name would imply. We only used the path portion, so passing the full
URL did lead to complications.
Now just the `pathname` is passed in, rather than the full URL. This
simplifies the metrics logic, and it incidentally fixes two bugs.
The main bug fixed is regarding Firefox metrics. Previously we had
assumed the `currentPath` would start with `chrome-extension://`, which
of course was not true on Firefox. This lead to us incorrectly parsing
the `currentPath`, so path tracking was broken for Firefox events.
This broken parsing is now bypassed entirely, so metrics should now
work the same on Firefox as on Chrome.
The second bug was that we were incorrectly setting the tracking URL
for background events during tests. As a result, we were incorrectly
detecting ourselves as an internal site that had referred the user to
us. But this was not of major concern, since it only affected test
metrics (which get sent to the development Matomo project).
Lastly, this change let us discard the `pathname` parameter used in
the `overrides` parameter of the `metricsEvent` function. Now that
`currentPath` is equivalent to `pathname`, the `pathname` parameter is
redundant.
* Remove `url` parameter from `metricsEvent`
The `url` parameter was used to override the `currentPath`, but it
never worked correctly. It was supposed to be used for setting the
`url` query parameter that was sent to Matomo, but `currentPath` was
always used even if it `url` was set and `currentPath` was empty.
Instead, `currentPath` is now always used. There was never a need to
provide an "override" for `currentPath` when it can be set directly.
The metrics provider does set `currentPath` automatically by default,
but this can be overwritten already by passing a second parameter to
`metricsEvent`.
There were two places this `url` parameter was being used: background
events, and path changes. Background events were submitted with no
`currentPath`, so because of the bug with the `url` parameter, the
metrics utility would crash upon each event. So those were never
actually sent. This commit will fix that crash.
The `currentPath` parameter was supplied as an empty string for the
path change events, so those never crashed. They just had the `url`
query string parameter set incorrectly (to an empty string). It should
now be correctly populated, which should mean we'll be capturing all
path changes now. Previously we were only capturing path changes to
pages that happened to include an event, because of this blank `url`
problem.
* Use `url` query parameter as fallback for generating `pv_id`
The `pv_id` parameter currently isn't generated correctly on Firefox,
as the generation assumes that the current URL starts with
`chrome-extension://`. The `url` query parameter is still unique for
each path, so it's probably good enough for generating an id for each
page.
This is just a temporary fix; it will be removed in a future PR, where
Firefox will be properly supported.
Background events are now sent in the `Background` category, rather
than `backend`. Conventionally we use the term "background" over
"backend", as it's not really a "backend" in the normal sense since
it's a client background process. Also it's capitalized because all of
the other event categories are capitalized as well.
The metrics URL has been updated to use `background` instead of
`backend` as well, for consistency.
Luckily we don't have to worry about our metrics being disjointed due
to this name change, because the background metrics never worked to
begin with! So there will be none under the old name. The metrics will
be made functional in a separate PR.
The Sentry DSN is now expected to be provided via environment variable
for production builds. The build script will fail if it is missing, and
an error will be thrown at runtime if it is missing.
The `SENTRY_DSN` environment variable has been set in CI to the old
value for `SENTRY_PROD_DSN`. We can migrate to a new DSN at some point
in the future.
In a non-production environment, Sentry was configured to send error
reports to a "test" MetaMask project. It will still do this during e2e
tests, but in development Sentry is now disabled completely.
In practice this was never useful in development.
* Fix popup/notification when browser is in fullscreen, primarily on OSX.
The issue was reported internally via Slack. User was running Mac OSX Chrome in fullscreen mode where Chrome is created in a new Desktop workspace.
The issue reproduced on OSX Chrome in fullscreen/maximized view overrides the explicitly set width and height for `windows.create()`. Possibly not overrides, but creates a window based off of the window that it was created from. Found a related [Chromium bug](https://bugs.chromium.org/p/chromium/issues/detail?id=263092&q=window%20create%20width%20os%3DMac&can=2).
The fullscreen `popup.left` pixel will calculate the window position incorrectly since we set and assume the width of the created window. The incorrect `left` position the window and transition the focus Desktop/Workspace incorrectly and make is seem to lose focus of the new window/workspace. Incidentally this will make the popup full width/height, and create a new workspace for the view, which we have no control over until Chrome
fixes it.
This will check if the popup is 'fullscreen', which it gets passed from the origin window, if so then don't reposition the window. If Chrome fixes the issue we can revert this change.
* Feedback commit
Co-authored-by: Mark Stacey <markjstacey@gmail.com>
Co-authored-by: Mark Stacey <markjstacey@gmail.com>
The `extra` property of errors sent to Sentry is sometimes not
initialized when we add the application state. A check has been added
to initialize it if it's missing.
I suspect that this changed with v5 of `@sentry/browser`, though I
can't find any explicit confirmation of this in their changelog.
The state snapshot that was attached to Sentry errors was removed
recently in #8794 because it had become too large. The snapshot has
now been restored and reduced in size.
A utility function has been written to reduce the state object to just
the requested properties. This seemed safer than filtering out state
that is known to be large or to contain identifiable information.
This is not a great solution, as now knowledge about the state shape
resides in this large constant, but it will suffice for now. I am
hopeful that we can decorate our controllers with this metadata in the
future instead, as part of the upcoming background controller refactor.
A separate `getSentryState` global function has been added to get the
reduced state, so that the old `getCleanAppState` function that we used
to use could remain unchanged. It's still useful to get that full state
copy while debugging, and in e2e tests.
An optimization in `account-tracker.js` was being skipped consistently
due to a type error (a number was being compared to a string).
The optimization in this case was to update the balances for all
accounts with a single request, rather than one request per account.
There were three cases where execution unintentionally continued after
an error was encountered. These cases likely are impossible to
encounter in practice due to recent validation improvements in the
`eth-json-rpc-middleware/wallet` module, but they were broken
nonetheless.
Execution inside the Promise constructor now halts immediately after
`reject` is called.
The state snapshot we were attaching to Sentry errors was too large.
As a temporary solution, it has been removed completely. We can re-add
it later after reducing its size.
Mark Stacey
Whymarrh Whitby
Thomas Huang
Erik Marks