All notable changes to this project will be documented in this file. See Conventional Commits for commit guidelines.
Custom attributes may now optionally be persisted between page loads via session storage. The API method setCustomAttribute now takes a third parameter, which defaults to false. When true, this boolean indicates that the provided custom attribute should be persisted in session storage and restored on subsequent page loads. Session storage is subject to the same-origin policy of the browser.
A new setUserId API method is now available. It defines a custom attribute containing a specified user ID, to be included in all payloads. The setUserId API is an extension of the setCustomAttribute API and automatically persists between page loads.
The previous npm package release included code that was not compatible with browsers in our supported browser list. The package exports will now default to code that matches our browser support list. The package will also include the source code for power users that would like more control over how the agent is built.
Some dangling, unused, and non-existent imports and exports were causing issues with consuming the NPM package in some projects. These have been removed.
Fixed an issue where typescript projects attempting to import the browser agent library would not get proper auto-complete in the import statement and would receive module declaration errors. The browser agent will not use tsc to generate type definitions from the JavaScript code at build time. The package.json has been updated with appropriate type export declarations.
Added error catch logic around the sendBeacon usage. In the event sendBeacon throws an error during the final harvest, the final harvest will attempt to fall back to the image data submission method.
Removed the array reduce polyfill in favor of the native browser array reduce method.
The agent will now utilize semantic versioning for subsequent releases. The previous version (1228) will carry forward into 1.229.0, and so on.
In an effort to align and standardize the timings the agent collects, the agent now uses the Google CWV library to track page timings such as CLS, LCP, FCP, TTFB and more. See Core Web Vitals documentation for more information.
The Browser Agent repository will now be available as a pre-release NPM package. See @newrelic/browser-agent for more information.
Fix an issue that caused session trace offset timings to be miscalculated in the early-page lifecycle, sometimes leading to negative "backend" timings.
The interaction-to-next-paint metric is now calculated and reported at the end of user sessions, via the Google CWV library. In addition, long continuously executed and blocking scripts detected by the PerformanceLongTaskTiming API are also forwarded to New Relic. This latter functionality is off by default, until a curated UI experience is created to utilize this data.
Partial revert of graceful handling change made in v1225 that unwrapped modified global APIs and handlers, which caused integration issues with other wrapping libraries and code.
Internal metrics were added to track the feasibility and impact of collecting page resource information using the PerformanceObserver resource timings, such as scripts, images, network calls, and more.
Added resiliency code around SPA interaction node save functionality to ensure a cancelled interaction node without a parent further up the interaction tree does not cause an exception to be raised from the agent.
Collate all of the internal statistic metrics calls, which--of today--are sent at page start and periodically, into one call made when the end user is leaving the page.
Customers were losing visibility into all calls on the page when denying timeslice metrics based on the deny list. This change reverts to the behaviour seen in all previous versions of the Browser Agent.
Updating the agent default configuration to enable the back/forward cache feature previously released in version 1222 by default.
The agent will attempt to handle niche objects throw from unhandledPromiseRejection events more gracefully. These cases could include objects with frozen or static properties, or custom extensions of the Error class without a set method in place.
Fixing issue where metrics harvesting was not being halted when the agent RUM call indicated the account did not have entitlement to the jserrors endpoint. Before this change, customers missing this entitlement would see network calls to the New Relic jserrors endpoint result in 403 or 409 errors.
This change forces all agents to call ingest at runtime to ensure that entities can be synthesized in NR1. This particularly pertains to any bespoke agent builds that did not utilize the page_view_event feature.
When importing of agent code fails, as when content gets blocked, the agent will clean up memory, the global scope, and any wrapped or modified APIs as much as feasible.
The agent's wrapping of the Promise object has been refactored to address conflicts with third party libraries and to add newer methods available on the native object. The new wrapping implementation is more conventional and less error-prone.
In some cases of failure to import agent script chunk "629", an error was thrown rather than caught and logged as a warning. The uncaught promise error responsible for this unintended behavior has been fixed.
In previous versions, the agent script included relative paths to its lazy-loaded chunks, which Googlebot picked up and attempted to index as pages. This change removes those relatives paths from the loader and centralizes our lazy chunk loading of agent features.
Previously, the agent didn't collect SPA browser interactions, XHR events, or session trace data in Chrome for iOS, which uses the webkit engine with modifications. The agent now collects the same data in Chrome for iOS as in other supported browsers.
Fixed an issue where multiple custom interactions harvested at the same time would result in only one interaction being persisted in New Relic.
Prevent time slice metric collection for AJAX calls when such a call matches an entry in the AJAX deny list.
Some browser versions will throw errors if sendBeacon doesn't have the navigator scope bound to it. A fail-safe action of binding the navigator scope to sendBeacon was added to try to support those browsers.
The build version is exposed to the newrelic global object. You can access it with newrelic.intializedAgents[<agentID>].runtime.version.
A new release of the browser agent will automatically raise a PR to the documentation site with the relevant changelog items.
The agent will attempt to preserve unhandledPromiseRejection reasons as human-readable messages on the Error payload that gets harvested. The previous strategy didn't always work, because Promise.reject can pass any value, not just strings.
Fixed an issue where when using the SPA loader with Next/React, route changes that lazy loaded components wouldn't be captured. While the issue specifically called out Next/React, this should apply to Nuxt/Vue and Angular.
Fixed an issue where when using the SPA loader with Angular, route changes that contained API calls, via Angular resolver, wouldn't capture the xhr/fetch on the interaction. This works with eager and lazy routes in an Angular SPA.
This architectural release simplifies file structure and refactors the way features are composed, in preparation for future developer experience improvements. These changes are not anticipated to have impact on agent behavior or functionality.
An instrumented page's back-forward cache eligibility was hampered by the agent's unload listener, which will be removed when a feature flag is on. With the allow_bfcache enabled in the init config, the agent's definition of (the end of) an user's session is more refined, and it will no longer be blocking the browser from utilizing its respective b/f cache.
The agent will now attempt to shut down an initialized feature if account entitlements are invalid. Accounts that lack entitlements to use certain endpoints will see many 403 errors in the console without this behavior. This behavior requires the Page View Event feature to be enabled.
AJAX events for data URLs have not historically been collected due to errors in the agent when handling URLs without hostnames. Going forward, XHR calls to data URLs will not cause agent errors and will continue to be excluded from collection.
The agent is now compatible with only modern web syntax (ES6+); this reduces loader size for these browsers by 20% or more. We target and test support for just the last ten versions of Chrome, Edge, Safari, and Firefox -- see browser agent EOL policy for more details.
Restoring previous functionality whereby the nrWrapper agent method should be excluded from JavaScript error stack traces.
Fixing an issue where external code redefining the self global variable causes the agent async loading to fail and the agent to crash.
Ensure the agent does not crash when sessionStorage is not available or when the quota has been exceeded or set to 0. Safari has been known to set the sessionStorage quota to 0 in private browsing windows.
The agent's infrastructure will now allow for the agent to be built to run on web workers for future projects.
To address "mismatched anonymous define" errors thrown by RequireJS, the agent's webpack library output will no longer include UMD checks for CommonJS and AMD module environments, and will instead be exposed globally via self.
Fixed an issue where custom attributes could be cleared and reset if the info block was included on the page below the loader script. Our guidance still remains that all configurations should be included on the page above the loader code, however this is an attempt to do no harm when feasible for backwards compatibility.
The Agent will now take into account the error object type, message, and original stack trace when deciding on whether multiple JS errors should be bucketed together.
PRs will run an action to detect workflow changes for a warning layer against vulnerability.
Fixed an issue where when using the SPA loader with Nuxt, the initial page load interaction was never being completed. This resulted in events like errors being retained in memory and never harvested because they were tied to an incomplete interaction.
Fixed an issue with the jsPDF library where it was not correctly detecting browser native support for Promises due to our wrapper. This resulted in an exception and jsPDF not generating the PDF. This issue is not present with the pro or lite agent.
Note: This issue does not affect the pro or lite agent. This change allows the jsPDF library to function correctly when the spa agent is used. However, it does cause an internal error within the agent to be generated. This error does not break the agent, jsPDF, or other functionality. The issue is planned to be addressed in a future update.
Pull requests will now generate and ship a build which gets consumed by NR1 dev components
- Internal NR Platform release date: 10/5/2022
- Production APM-injected release date: 10/6/2022
- Production Standalone release date: TBD
The Agent will now observes and captures unhandled Promise rejections as JavaScript Error objects.
Certain dependencies were appending non-ASCII characters to build files. These characters were affecting older Python agent implementations downstream that worked to encode the agent snippet. The build files are now checked and cleaned of non-ASCII characters before shipping.
The browser agent no longer uses 3rd party cookies to maintain and track session information. 1st party implementation using window.sessionStorage is now used, which is automatically cleared when a page session ends.
LCP is no longer reported on initialliy hidden pages
LCP metrics are no longer reported on pages whose state is hidden at load time, such as tabs refreshing in the background of a focused tab.
Individual features of the browser agent can now be dynamically loaded, enabled, or disabled at runtime.
The agent no longer inserts other features into the page via a script tag insertion. It now uses network requests to instantiate other code modules.
In an effort to better support the majority of our traffic, the test suite required to merge PRs has been updated to run against the latest 10 major versions of Chrome, Firefox, Edge, Android, and the latest 5 major versions of Safari and iOS. As part of this process, outdated code and polyfill libraries aimed at supporting deprecated browsers are no longer included in production builds by default.
Polyfills for IE11 have been included with the agent bundle.
BrowserInteractions no longer generate circular trees when they are self referential
Note: Version bump only for package newrelic
- Internal NR Platform release date: 4/19/2022
- Production APM-injected release date: 4/20/2022
- Production Standalone release date: 4/27/2022
Added internal mechanism for applying regex and replacement rules to all strings in payloads to obfuscate before sending to ingest.
A change has been implemented in our handling of applications hosted locally on a file:// protocol. For security reasons, we can not send payloads that contain file information, so our previous implementation revolved around completely shutting the agent down when file:// protocols were detected. This change introduces the ability to obfuscate all file:// paths from our payloads before sending to ingest.
The final character in trace ID hex generation was returning as undefined, which translated to always be 0 (undefined & 15 === 0). This change fixes this final character and ensures it is valid.
- Internal NR Platform release date: 01/24/2021
- Production APM-injected release date: 01/25/2021
- Production Standalone release date: 01/31/2021
Added front end framework detection metrics to help guide future priorities for browser agent features. The following front end frameworks will now be detected and analyzed:
- React
- Angular
- AngularJS
- Backbone
- Ember
- Vue
- Meteor
- Zepto
- Jquery
- Internal NR Platform release date: TBD
- Production APM-injected release date: TBD
- Production Standalone release date: TBD
Previously, XMLHttpRequest and Fetch calls made with Data URLs could prevent the agent from harvesting data.
Renamed LargestContentfulPaint PageViewTiming attributes from url to elUrl and tag to elTag. This makes the names less generic and as a result less likely to collide with custom attributes.
- Internal NR Platform release date: 12/06/2021
- Production APM-injected release date: n/a
- Production Standalone release date: n/a
The agent will now include core web vitals page view timings in the session trace waterfall payload. If observed, events such FI, FID, LCP, FP, FCP etc. will now be available in the browser Session Traces UI grouped under the timing type.
If a session trace ID exists, it will now be appended to harvests for the linking of session-related datasets downstream. This ID will be appended to any payload that exists at the time the session ID is determined, meaning the only harvests which won't have a session ID are PageView and Initial Page Load Browser Interaction.
The core web vitals metrics LCP and FI will now include metadata describing the network information observed on the page. This includes network type, round trip time (rtt) and downlink.
LCP metrics will now also report a tag name and an image URL if present (for LCP triggered by images).
- Staging release date: 11/04/2021
- Production APM-injected release date: 11/08/2021
- Production Standalone release date: 11/16/2021
Largest Contentful Paint will now stop being tracked when page visibility changes to hidden. This aligns with the web-vitals library.
Using the addEventEventListener without the passive flag for the touchstart event is flagged in Lighthouse. The passive flag is now applied to all addEventListener calls in the agent.
Array.isArray() call is not supported on old browsers and can cause a runtime error. This call has been replaced with an alternative that works on older browsers.
When a null value was passed in to the addEventListener API, the agent would report an internal error. This edge case is now being handled.
There was an edge case where certain ajax calls could have been excluded when they should not have been. This is now handled correctly.
The agent can now send metrics that capture information about how the agent itself is working internally.
- Staging release date: 09/27/2021
- Production APM-injected release date: 09/29/2021
- Production Standalone release date: 10/21/2021
Previously, XHR/fetch requests were captured as AjaxRequest events only when they were part of a route change. With this change, all requests will be captured as events. This feature can be further configured by specifying which requests should not be collected.
Previously, Span events were generated only for XHR/fetch requests that were part of a route change. With this change, all requests will be captured as Spans.
The CLS calculation has been updated to use session windows in order to align with Google Chrome tooling (Lighthouse, PageSpeed Insights, CrUX). For more information, see this blog post on web.dev.
The agent no longer calls the clearResourceTimings API, which had the potential to affect other scripts from accessing all resources. Instead, it now uses the PerformanceObserver API to collect information about resources.
- Staging release date: 07/01/2021
- Production APM-injected release date: 07/06/2021
- Production Standalone release date: 07/19/2021
PageHide PageViewTiming events are used to query CLS values. In cases where the page was never hidden, inconsistencies would arise because the PageViewTiming event with that type would not be collected. Now when pageUnload fires, if a pageHide PageViewTiming has not already been set, it will set it to the time of unload.
Perfect CLS scores were being ignored, because a score was only recorded when content shifted. This change reports perfect scores as 0, fixing inconsistent CLS queries.
Fetch calls are currently only recorded as AjaxRequest events with SPA browser interactions. This change records fetch calls as AJAX metrics, which will make them visible in the AJAX UI charts.
- Staging release date: 05/24/2021
- Production APM-injected release date: 05/26/2021
- Production Standalone release date: 6/2/2021
Up to 120 PageAction events can be harvested every 30 seconds.
The final Session Trace node in a harvest, captured using the Resource Timing API, is no longer duplicated in the subsequent harvest.
This issue lead to 1 duplicate node in a Session Trace, every 10 seconds, over the duration of the trace.
Fixed a memory leak in the agent when the network request to load the second part of the agent is blocked.
Fixed an error thrown in the console when the agent is loaded using the file:// protocol caused by features in the agent trying to run when others had been aborted.
Removed a legacy behavior used to ensure network connection was kept alive in IE 7/8/9.
Fixed an issue where route change Browser Interactions would wait forever if a setTimeout was called without a callback function (passing code in as a string in the first argument) as the first argument.
Fixed a conflict between the Browser agent and the Cypress.io test framework when instrumenting XMLHttpRequest.
- Staging release date: 03/10/2021
- Production APM-injected release date: 03/11/2021
- Production Standalone release date: 03/22/2021
The agent retries harvest XHR requests when it receives 408, 429, 500 or 503 response codes.
The agent will not report any data when it is on a page opened from a local file.
- Staging release date: 01/29/2021
- Production APM-injected release date: 02/01/2021
- Production Standalone release date: 02/08/2021
The agent now sends JS errors and AJAX metrics data as body of a standard XHR request.
- Staging release date: 01/07/2021
- Production APM-injected release date: 01/11/2021
- Production Standalone release date: 01/19/2021
The promise instrumentation has been updated to reduce performance overhead on web sites that use a large number of promises.
In a rare case where large number of callbacks are executed at the end of an interaction, the agent could cause a significant overhead. This has been fixed in this version of the agent.
Added handling for the use of fetch with a URL object.
- Staging release date: 08/18/2020
- Production APM-injected release date: 08/19/2020
- Production Standalone release date: 08/26/2020
The agent now reports two additional types of timing values as PageViewTiming events - pageHide and windowLoad. The pageHide value represents the first time that the page was hidden (e.g. by switching browser tab). Note that we only collect the first pageHide events at this point. This timing is useful alongside with its CLS (Cumulative Layout Shift) attribute.
In addition, the agent now also collects the legacy window load value as a PageViewTiming event. This is useful to query and visualize this metric alongside other types of timing events.
The agent now collects CLS (Cumulative Layout Shift) values as attributes on PageViewTiming events. CLS measures how much layout of the page shifts and is represented as a score. All types of PageViewTiming events (except FP and FCP) include this attribute, showing the score up until the point the timing measurement was taken.
Older browsers report Event.timeStamp as an epoch time instead of value relative to the page navigation start. The agent took this into account for FID (First Input Delay) timing values but not for FI (First Interaction), which goes hand in hand with FID. With this fix, there should no longer be unrealistic outlier values for FI values.
- Staging release date: 07/23/2020
- Production APM-injected release date: 07/28/2020
- Production Standalone release date: 08/03/2020
The agent can now use the W3C TraceContext headers in addition to and instead of the newrelic proprietary header.
- Staging release date: 05/20/2020
- Production APM-injected release date: 05/28/2020
- Production Standalone release date: 06/01/2020
The agent now accepts new configuration privacy.cookies_enabled. When it is set to false, the agent does not write any cookies, and it also notifies the intake server to not return a cookie. This enables customers to comply with GDPR privacy rules around cookies.
In some cases, the agent was causing a DOMException error when getting size of XHR responses for requests with document response type.
- Staging release date: 02/07/2020
- Production APM-injected release date: 02/07/2020
- Production Standalone release date: 02/07/2020
Resolved a bug that caused a script error when the windowUnload event fired. This issue affected only Internet Explorer prior to version 9.
- Staging release date: 02/03/2020
The agent is now capturing Largest Contentful Paint (LCP) as a new type of the PageViewTiming event. For more information about LCP, see this article from Google.
The agent is now capturing the timing of the Window unload event as a new type of the PageViewTiming event.
Requests with ms-stream data were previously causing errors in the agent.
- Staging release date: 12/18/2019
- Lite release date (US): 12/19/2019
- Lite release date (EU): 12/19/2019
- Pro / Enterprise release date: 12/19/2019
- Standalone/Current release date: 12/30/2019
The agent can now add the newrelic DT header to outgoing cross-origin AJAX calls. The origins that the agent should add headers to must be defined in the distributed_tracing.allowed_origins configuration section.
- Staging release date: 11/08/2019
- Lite release date (US): 11/14/2019
- Lite release date (EU): 11/19/2019
- Pro / Enterprise release date: 11/19/2019
- Standalone/Current release date: 11/21/2019
[Timing] New PageViewTiming Event to capture User Centric Perceived Performance metrics in real time
With this release, we are tying together the visual and responsiveness performance for our customer's site. With every page view, the agent is now capturing the time of the first interaction along with FID (First Input Delay). The existing paint timing metrics (First Paint and First Contentful Paint) are now collected even when they occur after the load event. All of these new metrics are captured on the new PageViewTiming events, available in Insights. This new event type is available for all agents, but requires a Browser Pro subscription.
The agent can now capture Span events for DT traces. This is accomplished by adding the custom newrelic header to outgoing AJAX calls. In this release, the agent adds this header only to same-origin requests. The feature must be enabled using the distributed_tracing configuration section.
- Staging release date: 11/01/2019
- Lite release date: 11/05/2019
With every page view, the agent is now capturing the time of the first interaction along with FID (First Input Delay). The existing paint timing metrics (FP and FCP) are now collected even when they occur after the load event. All of these new metrics are captured on the new PageViewTiming events in Insights.
Browser interactions measure the time of all Javascript code that runs as a result of an interaction. The agent now includes the time it takes to load and execute external scripts. This is useful, for example, when the code started by an interaction must be loaded first (lazy loading).
- Staging release date: 07/11/2019
- Lite release date: 07/16/2019
- Pro / Enterprise release date: 07/18/2019
- Standalone/Current release date: 07/25/2019
For browsers that implement the Paint Timing API, the agent will now collect paint timing values and make them available as attributes on the PageView and BrowserInteraction (initial-load only) events.
The history API methods are now instrumented on the History object constructor. This is to ensure that our instrumentation does not override other libraries that wrap these methods.
The DOM API methods used for JSONP instrumentation are now instrumented on the Node object prototype (as opposed to HTMLElement lower in the prototype chain). This is to ensure that our instrumentation does not override other libraries that wrap these methods.
- Staging release date: 04/17/2019
- Lite release date: 04/19/2019
- Pro / Enterprise release date: 04/19/2019
- Standalone/Current release date: 04/26/2019
Angular calls abort() on the XHR object after it successfully finishes. This was seen by our instrumentation as a call that did not finish, and as a result status code was set to 0. This fix addresses this use case by capturing status code earlier in the call stack.
- Staging release date: 01/02/2019
- Lite release date: 01/04/2019
- Pro / Enterprise release date: 01/08/2019
- Standalone/Current release date: 01/14/2019
This includes custom attributes added using the setCustomAttribute(), interaction.setAttribute(), and noticeError() APIs.
Currently, we are using the referer header value from the RUM call for transaction naming and for URL attributes on Insights events. The agent also sends the URL value as a query parameter with the RUM call to get around HTTP header stripping. This update brings the query parameter value on a par with the HTTP header by capturing it at the time the RUM request is made (to account for redirects).
- Staging release date: 10/02/2018
- Lite release date: 10/04/2018
- Pro / Enterprise release date: 10/08/2018
- Standalone/Current release date: 10/17/2018
The agent now captures the text of the HTML element that was clicked when a browser interaction started. This value is stored as an attribute called actionText on the BrowserInteraction events.
There is also a new API actionText, which can be used to manually set the action text value.
Browsers can return false from sendBeacon call when it cannot be completed. The agent now detects it and falls back to a different method to ensure data is captured.
The stackHash value was not being properly calculated for global errors in Safari 10 and 11, causing incorrect grouping of errors across all browsers.
On certain versions of the Safari browser, calling fetch without any arguments is permitted. Other browsers, in contrast, do not allow this and throw an error. This also prevented the agent from working properly.
Previously, the agent cloned the response of a fetch call and read the response body in order to capture its size. In certain versions of the Safari browser this caused other clone calls to fail. As a result, the agent now only uses the content-length header, when available, to capture response size.
- Staging release date: 11/14/2017
- Lite release date: 11/28/2017
- Pro / Enterprise release date: 12/04/2017
- Standalone/Current release date: 12/08/2017
When a JS error occurs inside a browser interaction event, the error will now be
associated with the interaction via Insights attributes. The JavaScriptError Insights
event will have browserInteractionId and parentEventId attributes, and BrowserInteraction,
AjaxRequest and BrowserTiming events will have browserInteractionId, eventId, and parentEventId
attributes.
Previously, the agent would cause Safari browsers to lock up when JSONP requests returned large data. The agent no longer calculates JSONP response size. https://newrelic.atlassian.net/browse/JS-3486
- Staging release date: 09/27/2017
- Lite release date: 10/02/2017
- Pro / Enterprise release date: 10/04/2017
- Standalone/Current release date: 10/11/2017
Browser Interactions that include JSONP requests are now correctly tracked.
Previously, browser interactions that included JSONP were ended early and not
included in the Breakdowns tab.
When New Relic and Zone.js v0.8.13 were used together, the context (this) was not
being set correctly in addEventListener calls.
When XHR was not wrappable, the agent did not send data to the collector via XHR. We are now correctly sending the data if XHR exists on the page, even if it's not wrappable.
- Staging release date: 06/30/17
- Lite release date: 07/05/17
- Pro / Enterprise release date: 07/10/17 1:20PM Pacific
- Standalone/Current release date: 07/17/17 9:15AM Pacific
The agent verifies interactions are complete by setting and clearing multiple timers. Previously, the agent would make many unnecessary calls to clearTimeout, and will noow only clear timers when appropriate.
When the agent determine the event origin for Session Traces. In some libraries,
that use custom event wrappers, when the agent calls target on an event it can
throw an exception. The agent will now catch the exception when building the
Session Traces.
- Staging release date: 06/08/17
- Lite release date: 06/13/17
- Pro / Enterprise release date: 06/15/17
- Standalone/Current release date: 06/22/17
On a mobile Safari browser all XHR's are not wrappable, resulting in errors in our agent. This change will no longer instrument SPA on these devices.
When you call setTimeout with a string as the duration, browsers will cast this as a number, but the agent did not handle this correctly. We are now handling this case correctly.
We have discovered that MutationObserver in IE causes a memory leak, so the agent now will prefer setImmediate for IE, and use a resolved promise to schedule the wrapping in Edge (and other browsers that support promises).
Some libraries recursively set timers that left our interactions open. The agent now handles this by reducing the max time allowable to be included in the interaction.
- Staging release date: 03/07/17
- Lite release date: 03/09/17
- Pro / Enterprise release date: 03/13/17
- Standalone/Current release date: 03/20/17
Currently, we limit PageAction events to 120 per page load and only 20 events per harvest cycle, resulting in dropped data if a user sends more than 20 pageAction events in a single burst. Increasing the harvest time to 30 seconds also increases the event buffer, allowing users to send up to 60 events per harvest.
The agent uses the system clock to calculate some timings, and since the system clock can change
over the lifecycle of a page, the agent will occasionally report inaccurate or unexpectedly negative values.
Going forward, the agent will use perfomance.now(), which is a monotonically increasing clock that starts
from navigationStart. This change will result in more accurate timing for modern browsers.
add- and removeEventListener should function the same whether the agent is present or not. This change completes
the bug fix from v993.
- Staging release date: 01/06/17
- Lite release date: 01/10/17
- Pro / Enterprise release date: 01/13/17
- Standalone release date: 01/19/17
- Current release date: 01/19/17
The release api was renamed from addReleaseId to addRelease, however the arguments did not change.
Previously, when the agent would send data using an XHR, it would feature-check the withCredentials property and set it to true if it was available. However, older versions of IE do not allow modification of the withCredentials property on unsent XHRs. The agent now wraps the property assignment in a try/catch block to prevent errors.
- Staging release date: 11/18/16
- Lite release date: TBD
- Pro / Enterprise release date: TBD
- Standalone release date: TBD
- Current release date: TBD
A new API was added that allows the client to inform us what version of their JavaScript is currently being ran, so they can access a richer error feature by helping match up which source maps can be used.
Previously, the wrapped scroll event listener did not take advantage of an available performance optimization: the passive annotation. By including this annotation, our wrapped scroll listener will allow a non-blocking, smooth scrolling action.
- Staging release date: 10/27/16
- Lite release date: 11/1/16
- Pro / Enterprise release date: 11/3/16
- Standalone release date: 11/10/16
- Current release date: 11/10/16
A series of changes created an issue that caused the event-emitter to drop the backlog after 30 seconds.
- Staging release date: 10/4/16
- Lite release date: 10/5/16
- Pro / Enterprise release date: 10/10/16
- Standalone release date: TBD
- Current release date: TBD
Previously Objects implementing the EventListener interface which were registered for multiple events, could only be removed for the last event they were registered for.
- Staging release date: 9/22/16
- Lite release date: TBD
- Pro / Enterprise release date: TBD
- Standalone release date: TBD
- Current release date: TBD
Previously when the agent and zone.js were both included on a page, additional
event handlers would be triggered twice. For example, when event handlers were
added as properties, such as onreadystatechange, these handlers would be
triggered twice in some browsers. This issue has now been resolved.
Previously, when adding event handlers for events in iframes, the agent would attempt to wrap the provided callbacks. When the wrapping logic called the callback belonging to another frame, a permissions exception would be thrown. The agent will now only wrap callbacks created in the same frame.
Previously, the agent would continue to buffer events to be processed and harvested, even if the aggregator portion of the agent failed to load. The agent now will clear the buffers and stop emitting events if it detects a failure, or if the rum request has not completed within 30 seconds of the load event.
Previously the agent would mutate the attributes object passed to add page actions by adding the default and page attributes onto this object.
The agent now has an api method to set the current route name for the page.
This api can be used to set the previousRouteName and targetRouteName for
BrowserInteraction events.
Previously the agent would send rum data to the router without TLS if the request was initiated from an insecure page. Now the agent will always use TLS connection when transmitting data.
- Staging release date: 8/16/2016
- Lite release date: 8/18/2016
- Pro / Enterprise release date: 8/18/2016
- Standalone release date: 8/24/2016
- Current release date: 8/24/2016
This release adds a new setErrorHandler api to agent which allows applications to see the errors collected by the agent, and optionally ignore them.
- Staging release date: 8/10/2016
- Lite release date: 8/15/2016
- Pro / Enterprise release date: 8/17/2016
- Standalone release date: TBD
- Current release date: TBD
Browser interaction events are now triggered by keyboard events as part of creating the interaction.
We now use the sendBeacon native API only for page unload.
This api restricts the amount of data it can send, so we will use
xhr when it is available and save sendBeacon for unload events.
We started sending back traced callback durations with the browser timing events. This is to match the attributes of other events.
- Staging release date: 7/5/2016
- Lite release date: 7/6/2016
- Pro / Enterprise release date: 7/7/2016
- Standalone release date: TBD
- Current release date: TBD
Previously the browser interaction api was only available in the spa loader used by our beta customers. The stubs for the api are now available in all loaders to allow switching between loaders without worrying about calling unavailable apis.
Previously the browser interaction stubbed out a deprecated version of the interaction api that was used briefly by beta customers. In this release this deprecated api is being removed completely.
Previously the agent did not properly wrap the fetch api during browser interactions. The agent now correctly wraps fetch, and the the body getter methods on Request and Response objects. It also correctly clones the fetch body before it is used to insure the agent can correctly measure responseBodySize.
Previously, SPAs that used hash-based routing would need to use the API to get meaningful names for their route change data, because the agent would strip the fragment from a URL and save only the path. The agent now sends the hash part of the fragment along with the path by default.
The addition of queueTime and appTime provides application timing data for breakdown charts by passing through server-side timing attributes.
In Firefox 38, copying the toString method from the native Promise class throws an error. The agent now returns a String representation of the original promise function, rather than throwing an error.
- Staging release date: 6/3/16
- Lite release date: 6/6/16
- Pro / Enterprise release date: 6/10/16
- Standalone release date: TBD
- Current release date: TBD
SPA agents now send data using the latest schema, and will now send navTiming data for initial page loads, and more detailed data for ajax requests including status codes, and requests/response body sizes.
- Staging release date: 5/2/16
- Lite release date: 5/4/16
- Pro / Enterprise release date: 5/5/16
- Standalone release date: 6/6/16
- Current release date: 6/6/16
When sending data to the router, all requests will now include a new query parameter which will contain the current url of the page. Previously the consumer used the referer header to determine the url of the page data was being collected for. This caused issues for sites that set a referrer-policy meta tag.
All spa api methods are now attached to an interaction handle returned by calling newrelic.interaction(). This handle will be bound to the interaction that was active when it was first created. The goal of this refactor is to allow more usecases to be handled by the api, and to reduce confusion caused by not knowing when an interaction is active.
The newrelic.finished() API call now again correctly records a PageAction
event with an actionName of 'finished' when it is invoked.
Previously, calling newrelic.endInteraction prior to the dispatching of the
window load event would cause SPA interactions after the initial page load to
not be submitted. This has been fixed.
Previous versions of our SPA instrumentation would generate bogus
BrowserInteraction events with a trigger of popstate when the hash was
changed during the initial page load interaction in some browsers. This has been
fixed.
Previously, each call to setTimeout or setImmediate that was recorded as
part of an SPA interaction would be sent as a separate record as part of the
data submitted for the interaction. For applications with lots of calls to
setTimeout(..., 0), this would result in unnecessarily large data payloads
being sent to New Relic.
Callbacks passed to setTimeout or setImmediate will now have their callback
timings rolled into the callback timings of the parent tracer that they were
spawned by instead, reducing the size of the submitted data for each
interaction.
When a custom attribute with a value of 'undefined' was attached via the
setInteractionAttribute or setCustomAttribute APIs, any browser interactions
containing that attribute would fail to be serialized correctly, and would thus
not produce BrowserInteraction events. This has been fixed.
- Standalone release date: 5/4/16
- Current release date: 5/4/16
The way that the JS agent was previously wrapping window.addEventListener was incompatible with the wrapping approach used by zone.js, which could lead to breakage of Angular 2 applications, particularly with respect to popstate handling. This has been fixed.
Previously, the agent's instrumentation of callbacks assigned via the XHR onreadystatechange property could cause those callbacks to not fire in some circumstances. Among other things, this affected the firing of some callbacks passed to jQuery.ajax. This has been fixed, and our test coverage of this area improved.
Previous versions of the SPA loader would cause a JS error on page load when used with Angular 2.x applications, due to a conflict with the zone.js library which is a dependency of Angular 2.x. This has been fixed.
The agent currently measures the page load duration as the time between navigationStart and the window load event, which is often a poor proxy for load time as experienced by the user.
The agent now provides another measure of initial page load timing that includes time spent waiting on XHRs and timers that don't resolve until after the window load event, and should be a more accurate reflection of user-perceived wait time.
This new timing is captured in the 'duration' attribute of BrowserInteraction events with a category of 'Initial page load'. It is currently only availble when using the SPA loader variant.
When instrumenting single-page web applications, the agent relies on detecting changes to the URL that are made by updating window.location.hash, or by using the history API in order to determine whether a given interaction should be counted as a route change or not.
Previously, route changes accomplished through direct assignments to window.location.hash might not be captured correctly in IE and Edge, but this has now been fixed.
Version 885 introduced a regression wherein the agent would fail to report URLs on individual JS error records. This would mean that JS errors might have been assigned a URL based on the URL at the time they were submitted, rather than the time they were recorded. This has been fixed.
This release bumps the querypack schema version to bel.2 which adds support
for jsDuration for browser interactions. It also removes the children property
from attribute nodes, and removes the className property from elementData nodes.
v862 introduced a memory leak in the core event buffering machinery of the
agent, which has been fixed in this release. The JS agent loader buffers events
to be consumed when the aggregator loads, but these buffers were not being
correctly destroyed starting in v862 when events were buffered using the
internal handle mechanism.
During development of the SPA feature, we collected data for many different types of user interaction. Many of these interactions are redundant (ie: mouseup, mousedown, and click), and added unnecessary noise into the event data. We now collect data for a subset of user interactions: click, submit, and popstate.
Previously, the agent would not correctly time XHR callbacks that were set up
using jQuery.ajax. It would also fail to include these XHR callbacks in SPA
interactions. This has been fixed.
v862 included a bug in the quality of our data collection, resulting in recording all BrowserInteraction events with a duration of 0 seconds. The agent now records the correct duration, and had improved test coverage of expected data.
The previous SPA custom instrumentation API included two methods with
confusingly similar signatures for tracking asynchronous and synchronous work.
To reduce potential confusion, the two methods have been merged into a single,
more general method: newrelic.createTracer.
Previously, Dirac events would only be written for interactions that resulted in a route change, but that determination would be made by URL comparison in the Consumer after the fragment identifier had already been removed by the agent. Route changes that resulted in changes within the fragment only would be ignored by the Consumer. Now, the determination of a valid route change is made in the agent before the fragment is removed, and sent as a flag to the Consumer.
Previously, new non-finalized SPA-specific API methods were available in all loaders. Now, the SPA-specific API will only be available for applications that opt in to SPA instrumentation by using the SPA loader.
In Edge, some SPA interactions involving Promises might previously have ended prematurely. This has been fixed.
Previously, the noticeError API would only accept an Error object. Now, a user can send either an Error object or a String to the noticeError endpoint.
Added new methods for attaching custom attributes to traced interactions, ending interactions early, adding custom segments to interactions, and assigning custom names to interactions.
These APIs are not yet considered stable and may change before the SPA feature is released.
The agent now includes a new loader variant called 'cap', which adds experimental support for capturing clicks as Insights events. This loader variant is not selectable in the UI, and should currently only be used for testing purposes.
In browsers that support it, the agent will now use navigator.sendBeacon to
harvest buffered data on page unload, rather than using a dummy image tag. A
related, issue wherein the next page load could be delayed in Firefox if data
submission took a long time has been fixed.
In Safari 9, pages with unload event handlers are allowed into the WebKit page
cache, meaning that those unload handlers may never fire. Since the agent
previously relied upon an unload handler to submit data when navigating away
from a page, this meant that data submission upon navigating away from a page in
Safari 9 was unreliable. This has been fixed by submitting data from the
pagehide event handler instead when possible.
Previously, XHRs requested with responseType='json' would trigger the agent to parse and re-serialize the response in order to measure the size of the response body. This could be a performance issue with large JSON responses requested with responseType='json'. The agent will now instead use XHR progress events to measure response size in most browsers.
A deprecation warning caused by calling webkitClearResourceTimings in Chrome 46+
has been fixed (we now prefer the un-prefixed version of this function if
available).
An Access Denied error was thrown when users were running >= IE10 in <= IE9 compatibility mode. The error was caused by submitting session trace data from a browser that does not support CORS. The agent will now only attempt session trace collection from browsers with known, working CORS support.
This version of the agent adds a new 'spa' loader which contains initial support for tracking route changes within single-page web applications.
The 3rd-party pace.js and rollbar.js libraries previously would interfere with New Relic's instrumentation of XMLHttpRequests, causing the 'AJAX' section of the browser UI to be empty. Compatibility with both of these libraries has been fixed.
Previously, when application code overwrote the window.Event global, session traces would be missing entries for event and timer callbacks, and internal errors would be generated in the agent, leading to unnecessary CPU usage by the agent when instrumenting high-frequency events. This has been fixed.
When attempting to serialize information about JS errors containing circular references in their 'message' property, the agent would previously hang for a long period of time, and then eventually fail to report the error. This has been fixed.
Query string parameters on URLs included within JS error backtraces will now be removed before error information is transmitted to New Relic. In addition, backtrace frames that reference inline scripts will be reported as 'inline', rather than the URL of the HTML resource.
PageAction events are now harvested more quickly - every 10s, rather than every 60s.
Some JS libraries monkey-patch XMLHttpRequest in such a way that the async
parameter will end up as false rather than true by default if unspecified.
To work around this, the agent now specifies the value of the async flag
explicitly to ensure that its XHRs are asynchronous.