We are working on the new 7.0 that will include Browsertime 3.0. More info soon.
-
To collect the Chrome timeline you should now use --browsertime.chrome.timeline instead of the old --browsertime.chrome.collectTracingEvents
-
To collect Visual Metrics add --visualMetrics (instead of the old --speedIndex)
-
You can now choose for what kind of content you want to include the response bodies when you use Firefox: --browsertime.firefox.includeResponseBodies 'none', 'all', 'html'
- The screenshot is not passed as messages anymore to decrease the memory impact.
- The Chrome trace log is not passed as messages anymore to decrease the memory impact.
- Upgraded to Browsertime 2.5.0 with fixes for the HAR in Chrome 66 and fix with User Timing measurements. Thanks @knaos for reporting and finding the issue.
- Upgraded PageXray to 2.2.1 that fixed measuring the size of 304 responses #1963.
- If the WebPageTest location contained a space, we sent that space to Graphite. That deosn't work, so we now make the location a safe key #1958.
-
You can now choose if sitespeed.io will return an error exit code if your budget fails with --budget.suppressExitCode see #1934 and #1936
-
You can now set ACL for S3 uploads #1937. Thank you lbod for the PR.
-
If WebPageTest fails, we now catch those errors better #1942. Thank you Lorenzo Urbini for the PR!
-
Running WebPageTest without Browsertime made the pages.pug fail #1945.
-
Upgraded to Pug 2.0.1 fixing various Pug problems caused by Pug internal version problems, see pugjs/pug#2978
-
Bumped dependencies #1952.
-
Fix HTML output summary for User Timings within Timing Summary table.
-
Docker: Kill some left over processes when you start a new URL, thanks Vitaliy Honcharenko #1952. We will fix this in another way in coming Browsertime 3.0.
- Upgraded from pug 2.0.0-rc4 to pug 2.0.0
- New cleaner pre-compiled WebPageReplay in the WebPageReplay Docker container
- Updated to latest Browsertime 2.2.2, checkout the changelog
- We now show CPU stats for Chrome in the HTML and send it to Graphite if you configure Browsertime to collect it #1914.
- Adding --filterList as parameter made changing the metrics filter fail #1915.
- Changing the metrics filter didn't work since 6.x, there was an assumption about messages that was wrong. #1912.
- There's been a major slowdown between 5 -> 6 when you generated HTML as reported in #1901. This has been fixed by #1909 and made faster than in 5.x by #1910.
- Adding back the -preURL options that mystically has been removed from the CLI (--browsertime.preURL still worked though). Thanks @aerwin for reporting - #1904
- There are user timings that broke the HTML output see #1900
-
In last release we accidently changed to only send first view metrics (by default) per page when you are using WebPagetest. We changed that and now send metrics for both first and second view. Thanks @wolframkriesing for letting us now.
-
Better default metrics for WebPageTest data in data storage. We now collect more metrics than before, see #1871. Thank you Jean-Pierre Vincent for contributing with your better default values. Jean-Pierre has also contributed with better dashboards for WebPageTest.
-
Upgraded to VideoJS 6.6 with smoother progress bar #1864.
-
Browsertime and WebPageTest plugin now sends browsertime.setup or webpagetest.setup when they are in the setup phase, so other plugins know that they will run #1875.
-
If you run WebPageTest standalone (without Browsertime) you will now get the the domains section using data from WebPageTest #1876 and you will get annotations in Grafana #1884.
-
PageXray is now a standalone plugin (before it was bundled with the coach). This makes it easier to use PageXray on HAR files from other tools (WebPageTest at the moment). #1877.
-
PageXray is now xraying WebPageTest HAR files (if you run WebPageTest standalone). This will add the PageXray tab per URL/run + the toplist and the assets tab #1880.
-
Upgraded the Docker base container to Ubuntu 17.10, NodeJS 8.9.4 and the WebPageReplay container with Firefox 58.
-
Added filenames to the video when you combine two videos in combineVideos.sh
-
New version of the Coach that now knows if you include Facebook in your page.
-
Upgraded to Browsertime 2.1.4 with new bug fixes and newer Chromedriver.
-
Fixed the start script so that you on Ubuntu can run WebPageReplay in the Docker container for your Android phone.
-
Chrome user timings was empty in the HTML output from WebPageTest #1879.
- Upgraded to PageXray 2.0.2 and Coach 1.1.2 that fixes #1861. Redirect chains that redirected back to the main page caused out of memory.
- The Docker container was missing the node MAX_OLD_SPACE_SIZE switch (so you can increase memory for NodeJS) #1861.
- Screenshot URLs in the HAR file was hardcoded to png, see sitespeedio/compare#11. That made jpg image links broken in compare.sitespeed.io.
- Use Chromedriver 2.34
- Configure the page complete time when you use WebPageReplay. Add -e WAIT 5000 to wait 5000 ms.
- Upgraded to PageXray 2.0.1 that fixes the Chrome problem with URLs that includes a #.
-
Make it possible to stop runs from your command line in the new alpha WebPageReplay docker container
-
Fixed bug with configuring pageCompleteCheck (and probably other problems too) in the Docker container #1858.
- We finally fixed (we hope) the SigV4 problem on uploading to S3 see #1689
- Better check for when the page has finished loading when you run WebPageReplay (load event end + 2 s).
-
Let plugin register message types for budget #1828. With this you can add your plugin metrics to the budget.
-
Let plugins run async JavaScript in Browsertime #1841.
-
Use sharp to change the size of the screenshot and choose between png/jpg #1838
-
Updated to Chrome 63 in the Docker container.
- Crawling now works with Basic Auth #1845 and #1506.
- Fix broken metrics list #1850. Thank you https://github.com/suratovvlad for reporting.
- Remove the unused unminified CSS file from the result #1835
- Updated to Browsertime 2.0.1 with fixes for Android.
- Yesterdays fix broke the functionality of adding/removing plugins with a comma-separated list #1833.
- Fixed a bug adding/removing multiple plugins see #1831.
Before you upgrade, please read our upgrade guide.
-
Use Chartist to display visual progress and size/requests to make it easier for users #1659.
-
The HTML pages now works better on larger screens #1740.
-
We upgraded to use the official Graphite Docker container and using Graphite 1.X as default #1735.
-
Log the full URL to your result, makes it easy to map logs vs result #1744.
-
Make it easier do build plugins: Expose messageMaker in the context to plugins (so plugins easily can send messages in the queue) #1760. Expose filterRegistry in the context so plugins can register which metrics should be picked up by Graphite/InfluxDb etc #1761. Move core functionality to core folder #1762.
-
Running Docker adds
--videoand--speedIndexby default to make it easier for beginners. -
You can now create plugins that can generate HTML (per run or per page summary). #1784.
-
You can now override/add CSS from your plugin by sending message of the type html.css #1787
-
Major work on the documentation: https://www.sitespeed.io/
-
The Coach 1.0 with tweaked advice about Google Analytics and Google Tag Manager and more.
- The
--plugins.loadand--plugins.disableoptions are deprecated in favour of--plugins.addand--plugins.remove. The previous syntax was cumbersome to use since it allowed for multiple plugins to be separated by space. When using it before the url argument, e.g.
sitespeed.io -plugins.load foo http://sitespeed.iothe url would be treated as a plugin name, and the command would fail.
-
Update to PageXray 1.0. For 99% of the users this will not change anything but if you where sending assets timings to Graphite/InfluxDB (as we told you not to do, these you now get blocked, dns, connect, send, wait and receive instead of just the total time #1693.
-
We removed the generic DataCollector that collected data for each run and instead each plugin should collect the data it needs #1731. If you have written a plugin that collect it owns data you can just follow the old DataCollector structure and move the code you need to your plugin. Also #1767 is a follow up to remove DataCollector.
-
We now default to Graphite 1.x so if you send annotations to Graphite < 1.0 you need to configure arrayTags to false
--graphite.arrayTags false -
We now output only the version number (and not package and version number) on --version.
-
As a first step to make it possible for plugins to generate HTML, we removed the hooks and instead only communicates with messages see: #1732 #1758. We now have three messages sent by the queue: sitespeedio.setup - The first message on the queue. A plugin can pickup this message and communicate with other plugins (send pugs to the HTML plugin, send JavaScript to Browsertime etc). The next message is sitespeedio.summarize (old summarize) that tells the plugins that all URLs are analysed and you can now summarise the metrics. The last message is sitespeedio.render which tells the plugins to render content to disk. The HTML plugin pickup sitespeedio.render, render the HTML and then sends a html.finished message, that then other plugins can pickup.
-
We have moved the GPSI outside of sitespeed.io and you can find it here. To run in along with sitespeed.io you just follow the instructions how to add a plugin. We moved it outside of sitespeed.io to make the code base cleaner and with the hope that we can find a maintainer who can give it more love.
-
The StorageManager API has been updated. The following changes may break code written using the 5.x API:
createDataDir(directoryName)is nowcreateDirectory(...directoryNames)and takes any number of directory names (which will be joined together) as arguments.writeData(filename, data)has reversed the order of its arguments. It is nowwriteData(data, filename).writeHtml(filename, html)has reversed the order of its arguments. It is nowwriteHtml(html, filename).writeDataForUrl(data, filename, url, subDir)no longer has a fifth argument indicating whether output should be gzipped.writeHtmlForUrl(html, filename, url)no longer has a fourth argument indicating whether output should be gzipped.
Note that all compression functionality has been removed. If you need compressed output, your plugin should handle gzipping itself. See the
harstorerplugin for an example.
- Upgraded to Browsertime 1.9.4 with latest Chromedriver that fixes launching Chrome > 61
- Fixed custom metrics problem with WebPageTest #1737
- Fix issue where coach, pagexray and browsertime data on summary pages might contain just a subset of data for tests with urls from multiple domains.
- Avoid crash if Google PageSpeed Insights request fails (e.g. due to incorrect API key).
- When you run sitespeed.io using Docker we now always set no-sandbox to Chrome (so you don't need to do that yourself).
- Custom metrics in WebPageTest broke the HTML #1722
- Skip storing faulty toplists on disk when using the analysisstorer plugin. The lists is generated from the raw data, so if you need them yourself as JSON, you can generate them #1721
- Upgraded to latest shining Browsertime
- Rollbacked to Chrome 60 to fix the flicker that happens on emulated mobile and makes lastVisualChange happens later than it should #367.
- Better logs when using WebPageTest.
- More finetuning in Browsertime (1.8.1) to pickup right last visual change on emulated mobile for Chrome.
- Showing the timer as default in the video that was accidentally changed in latest release. Thanks https://github.com/kkopachev for reporting!
- Use load time as of the default metrics for data storage when you use WebPageTest #1704
- Upgraded Browsertime to 1.8.0 with all these changes
- You can now run WebPageTest without Browsertime #1700.
- Use SpeedIndex to decide if a WebPageTest run works instead of fullyLoaded #1699
- Show in the CLI that requestheaders, blocking domains and basic auth work in Firefox.
- Upgraded to Browsertime 1.6.1 with a newer version of VisualMetrics that hopefully fixes the sometimes 0 metrics for some sites. #1961
- Include firstParty info in the HAR (more info about this soon).
- Also slack the screenshot of the run #1658.
- Upgrading the Docker container to use Chrome stable 60 (instead of beta 60)
- Upgrading to Browsertime 1.6.0 that gives support for adding request headers, blocking domains and basic auth in Firefox.
- When one WebPageTest run failed, it could break collecting metrics, it seems to happen on sites with many requsts. We now catch the error. See #1685.
- Upgraded to Coach 0.36.0
- Upgrading node-sass to work when installing on Windows 10 #1671
- Upgrading to Browsertime 1.5.4 checkout https://github.com/sitespeedio/browsertime/blob/master/CHANGELOG.md#version-154-2017-07-19
- Always add the first asset to the size list #1676
- 5.4.2 failed with NodeJS 6.11.1 since our base Docker container didn't include npm. This is fixed now and we use 6.11.1.
- Docker container uses NodeJS 6.11.1
- Latest Chrome beta 60.0.3112.66 in Docker
- Set 6.11.1 as minimum engine.
-
Updated the Docker container to contain fonts for Hindi, Thai, Japanese, Chinese and Korean.
-
Updated to Browsertime 1.5.3 that includes a fix for faulty content types when getting the Chrome HAR file #1654
- Upgraded to PerfCascade 2.0.2
- You can now choose max number of items in the toplists. Do it with --html.topListSize. #1639.
- You can now get a list of largest and slowest third party assets #1613.
- Upgraded to latest Browsertime:
- Upgraded to Geckodriver 0.17.0 seems to fix #321.
- Upgraded Chromedriver 2.30 with a very special hack to fix #347.
- Pickup metrics from the Paint Timing API #344, will work in Chrome 60.
- Updated the Docker container to Firefox 54 and Chrome 60 (beta) to fix the background color problem. Chrome bug 727046.
- If you run Chrome 60+ you will now see the metrics from the Paint Timing API in the Browsertime tab.
- Fixes for custom filters for PageXray #1647
- Fixed bug when calculating timing for an assets included both SSL and Connect time, that is wrong since connect time includes SSL time. This only affected the time showed in slowest assets toplist.
- Collect assets that have 2XX status code instead of just strict 200. Thanks @vio for the PR.
- Added domInteractive and domContentLoaded to the waterfall graph #1632.
- Show GPSI score on summary, detailed summary and pages columns #1631.
- Upgraded to Browsertime 1.4.0:
- --browertime.preURLDelay (in ms) so you can choose how long time you want to wait until you hit the main URL after the pre URL.
- Fixed setting proxy using --proxy.http and --proxy.https #338
- Updated to chrome-har 0.2.1 that: add "serverIPAddress" field to entries, set bodySize for requests correctly, set bodySize and compression for responses correctly, and add _transferSize field for responses, just like Chrome does.
- The link in the HTML to the Chrome trace log is not working.
- Upgraded to Browsertime 1.2.7 that downgrades Chromedriver to 2.28 to make collecting trace logs work again.
- Upgraded to Grafana 4.3.1 in the Docker compose file.
- Too many runs broke the HTML #1621.
- When you used --summary and --summaryDetail it broke the run (thanks gamerlv for reporting) #1622.
- Added support for --s3.storageClass option (thanks shakey2k2) #1623.
- Show browser version on HTML run pages (finally!).
- Upgraded to WebPageTest API 0.3.5, VideoJS 6.1, PUG 2.0.0-rc1 and PerfCascade 2.0.1.
- Upgraded to Browsertime 1.2.6 with a fix so that setting Firefox Preferences works as expected.
- Verify that you add the host when you want to send metrics to Graphite #1587
- Fix navigation making anchors miss target #1609. Thanks Radu Micu for the PR!
- Added a new script /tools/combineVideos.sh to combine two videos into one.
- Include the browser name in the file name when you download a file #1594
- Show backEndTime in Summary Timings (to make it easy to find runs with same backEndTime) #1595.
- Always log the WebPageTest id in the error logs if WebPageTest fails.
- Upgraded to PerfCascade 2.0.0 with performance improvements and keyboard accessibility.
- New Browsertime 1.2.5 with fine tuned video for Firefox, Basic Auth support in Chrome, fix for URLs containing commas and new version of VisualMetrics (thanks Pat) that can sort out the problem that started to happen with Chrome 58 running in emulated mode (or at least most of the times).
- Rework of the WebPageTest result tab: More metrics, requests per content type, requests per domain and screenshot.
- Rework of PageXray result tab: requests per content type, requests per domain, cookies stats
- Rework of Browsertime result tab: easier to find the most important metrics
- More info from GPSI on the result tab.
- Always download screenshots, waterfall graphs, and Chrome traceLog and store it local from WebPageTest.
- Better HTML page titles, showing what's tested and when, that makes it easier when you share URLs.
- HTML tuning for smaller devices.
- Added proxy parameters in the CLI from Browsertime.
- Updated to PerfCascade 1.4.0 (from 1.0.0)
- Add new s3.path option, to override the default storage path in the S3 bucket.
- Pickup timestamp from each run and display on each run page
- Added possibility to set the graphite web host (--graphite.webHost)
- Set Graphite tags as arrays (--graphite.arrayTags) needed for Graphite 1.0
- Add request headers with -r name:value (Chrome only in this release)
- Block domains with --block my.domain.com (Chrome only in this release)
- The default upload path in S3 buckets no longer includes the prefix 'sitespeed-result'
- Waterfall graphs loaded in Safari iOS didn't work.
- Allow S3 upload even when using custom outputFolder.
There's one change in 5.0 that changes the default behavior: TSProxy isn't default for connectivity since it doesn't work as we want together with Selenium. We also removed tc as default running Docker. When you change connectivity you should use our Docker network setup! If you used to use TSProxy or tc, please switch to Docker networks for now.
- Link User Timings in Page Summary (so you easy can find min/median/max of your User Timings).
- Slack missed URL specific information introduced in the 4.6.1 release. Now you get info per URL again.
- Default values for warning and error in Slack was wrong. Before warning was 80 and error 90, now warning is 90 and error 80.
- Setting a speedIndex/firstVisualChange as a warning/error value for Slack didn't work.
- Design: Make it easy to see which run you are looking at on by highlighting that run on the run list. Unified metrics naming.
- Made it possible to gzip the HAR files.
- Made it easy to download WebPageTest HAR files.
- A loader indicating that we are waiting in the HAR when it is fetched.
- New Chromedriver 2.28.0 that fixes "Cannot get automation extension from unknown error: page could not be found ..."
- The help for budget had wrong example parameter. Use --budget.configPath for path to the config.
- You can now choose to load the HAR file using the fetch API instead of inlining it in the HTML file. Use --html.fetchHARFiles #1484
- New version of Browsertime that has a new metric VisualComplete 85% (or more), thank you @jeroenvdb! You can see the metric in the Waterfall graph and it will automatically be sent to Graphite.
- Browsertime is now using https://github.com/sitespeedio/chrome-har to generate the HAR.
- Pickup the number of script tags on the page (from the Coach) and display it in the Coach section and send by default to Graphite.
- Show SpeedIndex, FirstVisualChange and LastVisualChange in colums for pages (so you can sort them).
- Send number of script tags, local storage size and number of cookies by default to Graphite.
- Link VisualMetrics from the page summary selection so you easily can go from Grafana to a specific run #1457
- Updated to Firefox 52.0 and stable Chrome 57 in the Docker container.
- Upgraded Grafana dashboards to use the latest metrics.
- Upgraded to PerfCascade 0.9.0 that is smarter when drawing time lines see PerfCascade #160
- Make sure we show preURL and connectivity type for all result pages #1493
- Making regions for S3 work #1486
- Annotation timestamp and metrics timestamps are now in sync #1497
- Rolling forward to Chrome 57 beta since 56 isn't working correct with our video, see #284
- Upgraded to PerfCascade 0.6.1 from 0.4.0 with UX fixes see https://github.com/micmro/PerfCascade/releases/.
- Added CLI options for setting Firefox preferences (--browsertime.firefox.preference), collect the response body in Firefox HAR (--browsertime.firefox.includeResponseBodies) and Chrome browser CLI args (--browsertime.chrome.args).
- You can now collect the timeline log from Chrome (--browsertime.chrome.dumpTraceCategoriesLog) and set which traceCategories Chrome should collect.
- If you collect the timeline from WebPageTest (--webpagetest.timeline) it will automatically be stored in your data folder and linked from that runs HTML page.
- New Browsertime release where you can set the connectivity type to external, that makes it possible to use Docker network bridges for setting up connectivity, more about that soon in the blog post. Thank you @worenga for higlighting the problem!
- New Browsertime that fixes a potential problem when generating the HAR for Chrome see BT #272
- Show graphite.auth and graphite.httpPort in the CLI to make it easier
- Make it possible to configure S3 uploads with: s3.maxAsyncS3, s3.s3RetryCount , s3.s3RetryDelay , s3.multipartUploadThreshold, s3.multipartUploadSize see https://www.npmjs.com/package/s3#create-a-client #1456
- Updated Docker container to use Chrome 56 and FF 51, but also added no-sandbox as default to make it work on Docker on OS X (new 56 thing).
- Updated to PerfCascade 0.3.7 where you open/close the extra request info by clicking on the bar.
- Use --webpagetest.script to supply your script as a string and --webpagetest.file as a file. But don't worry, it will work the same as before. Thank you Jeroen for the PR. See #1445
- Send the result HTML to S3 #1349
- Send annotations to Graphite with a link back to the HTML result #1434
- Surfacing user defined whitelist from browsertime for user timings filtering #1426
- Use connectivity native as default if no one is set in WebPageTest #1447
- Made it possible to set WebPageTest runs as non private #1448
- Fix for Template error with meta data aliases when not using the CLI #1444
- Worst case scenario if Browsertime missing a HAR file, the HTML summary rendering failed #1424
- If we have a site that is missing expire headers, the HTML generation failed 1430
- Updated to latest PerfCascade that will pickup changed resource prio in Chrome and some bug fixes.
- Google is still overloading User Timing marks see #257. This fix mute the marks from WebPageTest so they aren't sent to Graphite.
- Google is overloading User Timing marks see #257. This is quick fix, lets make a better fix in the future.
- New Browsertime that fixes the too early firstVisualRender in Firefox introduced in 4.3.5.
- Running only WebPageTest generated errors in the HTML plugin #1398, fixed in #1413
- New Browsertime (beta 22) with changed configuration for Chrome to detect orange frames
- New Coach and Browsertime that is cleaned up to make our Docker containers smaller again 726 -> 547 mb
- Upgraded to Browsertime beta 21 to finally once and for all fix the problem with the too early firstVisualChange that sometimes happend in Chrome:
- We removed the center cropping of images when visual metrics checks if an image is orange/white. The cropping made us miss the small orange lines that sometimes appear only in Chrome.
- We also fine tuned (and made configurable) the number when the diff of two images (orange and white) is ... orange.
- We re-arranged how we record the screen to record as little extra video as possible
- Upgraded browsertime with changed FFMPeg config to hopefully fix the too early firstVisualChange that sometimes happens for Chrome, updated Geckodriver (0.12.0), changed Firefox default settings to follow the Mozilla teams default ones when they do test automation, pickup changed request prio in Chrome (before only initial prio was used) and adding new connectivity profile 3gem for Emerging markets to keep in sync with WebPageTest.
- Updated the Docker container to use ImageMagick 6.9.7-2 to fix firstVisualChange that sometimes was picked up to early sitespeedio/browsertime#247
- TAP and JUnit XML stopped working when changing to yargs 6.x (coerce has breaking changes). Rollback to yargs 5.0 and make a better fix after the holidays.
- Upgraded to Browsertime beta-19 that fixes firstVisualChange happen to early when testing as mobile
- New version of PerfCascade with icons for content types, making it much easier to understand the waterfall graph.
- Custom metrics in from WebPageTest introduced a error running WebPageTest without custom metrics. #1389
- Updated to browsertime beta18 with fix for to early firstVisualChange with preURL and display lastVisualChange in the video. And fixed the extra 5s added to base request using Firefox in Docker. And makes the order for assets more correct in Chrome for sites using HTTP/2
- Upgraded the Docker container to use Chrome 55.0
- The metric lines (firstVisualChange etc) is now stronger and easier to spot
- Slack: firstVisualChange, lastVisualChange and fullyLoaded metrics will be sent by default and you can now configure what metric you wanna use to decide if it is a warning/error message #1366
- Use video.js as video player #1372
- Collect custom metrics in WebPageTest (and send them to Graphite when configured). Thank you https://github.com/jpvincent for the initial PR! #1377
- Added ability to set a custom alias for URLs via the text file for shortening long page URLs. #1326
- Trap to catch when you wanna exit a Docker run. Now you can just exit.
- Add support for pushing metrics to InfluxDB.
- Latest coach (0.30.0).
- Running multiple URLs in WebPageTest failed because of a "feature" in the WebPageTest NodeJS API where options in s are change to ms. #1367
- The keys for assets in PageXray was broken when we sent them to Graphite, because we couldn't identify which asset we sent, instead of the URL we used the position in the array. We fixed that now, BUT: Please don't send all the assets to Graphite, it will fill your disk! #1341
- The key summary structure for metrics for WebPageTest just worked because of luck. It is now divided in pageSummary and summary making it easier to configure and understand. #1377
- Fixed encoding problems when storing to disk #1346
- If you tested multiple runs, the video was overrun by the last URL, see sitespeedio/browsertime#237
- SpeedIndex, First/Last Visual change is now in the help section
- Show SpeedIndex, First/Last Visual change on the detailed summary page
- Show last visual change in the summary box
- Color the first/last change line in the waterfall graph
- Show legend for the waterfall graph
- Added breakdown of 1st vs 3rd party content types
- Color of tabs in waterfall graph is now white and readable. The URL in the tabs has the right letter spacing.
- Crash when all assets matched the specified first party regex. #1358
- Additional checks to avoid generating invalid paths in Graphite.
- New version of PerfCascade that gives us numbers on requests, image tab last and horizontel lines in subseconds.
- Output preURL info on the runInfo box on each HTML page.
- If we have first and last visual change add it to the HAR file so we can see it in the waterfall graph.
- Output Speed Index and First Visual Change in page summary box (the logic was there for SpeedIndex before but failed).
- Added missing shorthand --preURL to the CLI options.
- Cli help options for Browsertime was very unclear and unspecific.
- TSProxy is somehow broken together with Selenium. TC is now default connectivity engine when running in Docker.
- Finally fixed the problem with Chrome that it sometimes didn't start in Docker: SeleniumHQ/docker-selenium#87 (comment)
- Made the size table sortable for PageXray metrics
- Upgraded the Docker container to use FF 50
- Upgraded to latest Browsertime beta 13 with official video support
- Option to set your custom alias for connectivity thank you @jpvincent for the idea #1329
- GPSI now uses mobile configuration so if you pass --mobile, it will use the mobile rules. #1342
- Always send PerceptualSpeedIndex to Graphite as picked up by Browsertime/VisualMetrics
- Added --video and --speedIndex to record a video and get SpeedIndex and related metrics using VisualMetrics. Use it in our Docker container.
- If you configured to run a video you can see the video in the Browsertime tab.
- Upgraded to Browsertime 1.0.0-beta.11
- Fixed issue with incorrect values for speedindex and start render due to a small info bubble of text appearing in the video frames.
- Additional information added in the documentation around using connectivity engine
tcfor network throttling. - Additional information added in the FAQ section of the documentation mentioning Digital Ocean issue with pre-baked docker(1.12.3) instances and Firefox.
- Running budget with one rule for one URL failed the JUnit output, thanks @krukru for the report #1317
-
Pick up environment variables in the CLI. The namespace is SITESPEED_IO. This is useful for setting up default values in Docker. Say you want to set an environment variable for --browsertime.iterations 1 then use SITESPEED_IO_BROWSERTIME__ITERATIONS=1. Checkout https://www.npmjs.com/package/yargs#envprefix for full docs.
-
Upgraded to Browsertime 1.0.0-beta.10:
- Added initiator of each request entry to chrome HAR
- Output SpeedIndex & firstVisualChange in the logs if you use VisualMetrics
- Generating HAR files from Chrome caused a crash in some cases.
- Entry timings in HAR files from Chrome were strings instead of numbers.
- One extra fix for outputing timing metrics in the console: If timing metrics is < 1000 ms don't convert to seconds and let always have fixed
size for mdev fixing many numbers for SpeedIndex. - Configure proxies with --proxy.http and --proxy.https
- New TSProxy that is less complex
- Upgraded Selenium to 3.0.1 (no beta!)
- Upgraded Geckodriver to 0.11.1
- Updated minimum NodeJS to 6.9.0 (same as Selenium). IMPORTANT: Selenium 3.0.0 will not work on NodeJS 4.x so you need to update.
- Export chrome perflog dumps as json in extraJson property of the result, instead of a string in the extras property. Only relevant to api users.
- Upgraded sltc so we use 0.6.0 with simplified tc that actually works
- We now run xvfb from inside NodeJS so we can set the screen size, making it easy to record the correct size for VisualMetrics. We also use environment variables that starts with BROWSERTIME so we can turn on xvfb easily on Docker. https://github.com/sitespeedio/browsertime/blob/master/CHANGELOG.md
- If you where using graphite.includeQueryParams when you where sending keys to graphite, URLs containing ? and & failed. Those characters are now replaced.
- We rollbacked sending the URL in Graphite keys for toplists, it's opened the possibility to generate too much data in Graphite.
- FYI: Soon we will update to the final 3.0.0 of Selenium and we then need to drop support for NodeJS 4.x.
-
Setting --mobile didn't change viewport/useragent. Thank you @zhangzhaoaaa for reporting #1298
-
Asset toplist data (slowest/largest assets) couldn't be sent to Graphite. It's now restructured with new naming and you can also get the URL for slowest assets toplist in Graphite. Documentation coming up the coming days. #1294
- Domains metrics wasn't grouped per domain, making metrics sent to Graphite fail if you configure them to send all. See #1290 and #1289.
- If you configured cli params for Chrome, the check for Android configuration broke the run.
- The CLI output was wrong when setting up a custom profile. You should use --connectivity custom
Version 4.0 is a ground up rewrite for Node.js 6.9.1 and newer. It builds on all our experience since shipping 3.0 in December 2014, the first version to use Node.js.
-
We support HTTP/2! In 3.X we used PhantomJS and a modified version of YSlow to analyze best practice rules. We also had BrowserMobProxy in front of our browsers that made it impossible to collect metrics using H2. We now use the coach and Firefox/Chrome without a proxy. That makes it easier for us to adapt to browser changes and changes in best practices.
-
We got the feature that people asked about the most: Measure a page as a logged in user. Use --browsertime.preScript to run a selenium task to before the page is analyzed. Documentation is coming soon.
-
New HAR files rock! In the old version we use BrowserMobProxy as a proxy in front of the browser to collect the HAR. In the new version we collect the HAR directly from the browser. For Firefox we use the HAR export trigger and in Chrome we generates it from the performance log.
-
Stability: We have a new completely rewritten version of Browsertime that makes it easier for us to catch errors from the browser, drivers and environment problems.
-
Speed: Yep we dropped Java (it was needed for BrowserMobProxy) and most things are happening in parallel with the new version.
-
Don't overload Graphite: One thing that was annoying with 3.x was that it by default sent a massive amount of metrics to Graphite. That's cool in a way but it was too much. We now send curated metrics by default and you can choose to send more.
-
You can collect metrics from Chrome on an Android phone. In the current version you need to have it connected using USB to the server running sitespeed.io, lets see how we can make it better in the future.
-
Using our Docker container you will get support getting SpeedIndex and startRender using VisualMetrics. This is highly experimental at this stage.
And many many more changed. Read about the release https://www.sitespeed.io/sitespeed.io-4.0-is-released
- Call summary Site summary in the slack plugin to make it easier to understand.
- Run as root inside the Docker container, it makes things easier.
- You can now configure so sitespeed.io only slacks on error #1260
- Do not run as root inside the Docker container #1259
- Budget HTML page and log the budget info. #1264
- New browsertime beta-9 and latest Coach
- Sending all data to Slack changed property name from both to all: --slack.type all and default is now all (instead of summary).
- Don't set viewport when running on Android.
- Screenshots for Firefox works again #1256
- Cli for setting budget junit output is junit NOT junitxml
- Slack specific URL errors, so if something fails we know about it #1261
- Send summary and individual URL metrics to Slack. #1228
- Simple first step for performance budget #1227
- Download the HAR files to your local #1174
- Moved site to docs folder within the project #1240
- The data structure (internally) for toplists is changed so they can be sent to Graphite and used in a budget. Messages for largest assets was renamed from assets.aggregateSizePerContentType to assets.largest. Also send the largest individual size of an image to Graphite by default.
- Plugin analysisStorer is now called analysisstorer and messageLogger is now messagelogger and live within their own plugin folder, following the pattern of all other plugins.
- Default cli options now lives in each plugins.
- No default channel for Slack, use the one setup in the hook instead.
- Run sitespeed.io without the cli.
- A little better error handling when loading a plugin fails.
- LastModified in the PageXray section was wrongly outside the table
- Layout for resource hints
- Pug templates never was a cache hit, so generating the HTML took a lot of extra memory #1218 #1219 thank @moos for the PR #1220
- Fix crash for pages that didn't set the last-modified http header #1221
- WebPageTest data rendering was broken since 4.0.0-beta.1.
- You can now output a summary for a run in the CLI. Thanks @moos for the PR.
- The output directory structure was broken in beta.2, please update to beta.3.
- If the HAR export plugin fails in Firefox, don't break the run, use the data we have and try to make the most if it. #1190
- Make it possible to link to specific tabs on page/run pages #1087 Thank you @moos for the fix!
- Passing --outputFolder now works correctly. #1209
- Upgraded to latest Browsertime to make Firefox 49 to work again.
- Send content size & request per content type by default to Graphite for WebPageTest pageSummary #1194
- You can now set --max-old-space-size when running the Docker container, use -e MAX_OLD_SPACE_SIZE=4096 (or what max size you want) #1185
- The old hotlist is rebranded to toplist. You can now see the largest responses by content type and the slowest responses #1183
- Send content size & request per content type grouped by breakdown to Graphite for WebPageTest summary #1194
- Get plugin by relative or absolute path (perfect for Docker). If you wanna use a npm installed module use $(npm get prefix)/foo
- The CSS size advice on summary page used the wrong metrics to check the color, meaning 0 bytes made it red :/
- Shorten long URLs displayed on HTML Asset report #977
- Empty size fixes for tablet view of HTML report
- Title/description for each page says something about the page + we don't want robots to index the result pages
- Fixed long URLs that broke page design #1020 #1049
- Right CLI parameter name for setting the Selenium URL (if you use a standalone server)
- Made it easier to understand how to configure location and connectivity for WebPageTest.
- Fixed breaking GPSI when summary was missing from the rule. #1110
- Updated browsertime, so when you are running pre/post script, you can get hold of the selenium-webdriver from the context (a lot of people have had problem with it). Checkout https://github.com/sitespeedio/browsertime/blob/master/test/prepostscripts/preLoginExample.js for an example.
- Finally Firefox works again, you can use Firefox 48 since we now use Geckodriver.
- New version of TSProxy that fixes the redirect problems for HTTP -> HTTPS sitespeedio/browsertime#175
- New PerfCascade that takes care of responses that misses a content type #1030
- Always include the PerfCascade script (the path was wrong sometimes before) #1030
- Include PageXray metrics in the summary #1162
- Updated to latest Browsertime, now supporting different connectivity profiles using tsproxy or tc. #895
- Disable default plugins (disable HTML, screenshot etc). Looking forward to add plugins #1015
- Configure the pageCompleteCheck using CLI.
- Moved metrics to different tabs on the page result page.
- Set the number of runs for WebPageTest using a specific cli parameter #1101
- Run custom script and URL/script for WebPageTest #1101
- Run custom script. Use --browsertime.script myScript.js to add your script. You can run multiple script by passing the parameter multiple times. The metrics will automatically turn up on the summary page, detailed page, the summary page of the page and the run page. They will also be sent to Graphite. #1063
- Add your own plugin (examples coming soon) #891
- Removed analysisStorer as default plugin (do not store all json metrics by default). If you want to use it, enable it with --plugins.load analysisStorer
- Always show the waterfall if you run one run.
- You can configure which metrics to send to Graphite #943
- Send domain, expireStats and lastModifiedStats to Graphite for a group summary and totalDomains for page summary.
- Send documentHeight, domElements, domDepth.avg, domDepth.max and iframes per default for each page summary (missed those, these are handy!).
- All URLs is now belonging to a group. The group right now is the domain of the URL. When summaries are sent to Graphite, each summary belong to a group. Meaning summary metrics will always correspond to the group (not as before the filename if you get the URLs from a file). #1145
- New default namespace for Graphite metrics: sitespeed_io.default to be more flexible for the dashboards we will supply
- Removed p10 and p99 as default to Graphite (and all calculations). We make it configurable in the future from the cli (it's prepared already).
- It looks like url is field that can't be used in pug, so we where missing the URL on each individual page. Renamed and fixed.
- Added the detailed summary page that we learned to love.
- Small summary on each summary page to display the run information.
- More summary boxes on the summary page.
- Updated Browsertime to new release including latest Selenium to make Firefox 47.0.1 work.
- Added definition and help for the metrics in the summary boxes.
- Use PerfCascade to show waterfall on page summary and individual runs from collected HAR (#997)
- Show simple Google Page Speed Insights metrics in HTML and send the score to Graphite (#948)
-
Lets use Chrome as default browser since Firefox 47/Selenium/Gecko/Marionette is broken sitespeedio#993
-
Only add browser/connectivity in Graphite keys in metrics collected by Browsertime (#1009).
-
Add location/connectivity to Graphite keys for WebPageTest (#1008).
-
Internal: We changed from Jade to Pug (latest) and moved to Sass for the CSS (thanks @matthojo)
- Enable the coach by default.
- Fixed broken default plugin loading.
- Show results from other plugins even if one plugin has an error (e.g. WebPageTest fails but Browsertime succeeds).
- Fixed preTask and postTaks, now named preScript/postScript, try iy out to login the user.
- Added browser and connectivity to Graphite keys, all keys now hold browser and connectivity.
- Hey we now log the the log file in the result dir.
-
Everything! Rewrite from scratch in progress. This is an alpha release, try it test it but do not upgrade in production yet (sitespeedio#945).
-
We support HTTP/2! In 3.X we used PhantomJS and a modified version of YSlow to analyze best practice rules. We also had BrowserMobProxy in front of our browsers that made it impossible to collect metrics using H2. We now use the coach and Firefox/Chrome without a proxy. That makes it easier for us to adapt to browser changes and changes in best practices.
-
We now support the feature that people asked about the most: Measure a page as a logged in user. Use --browsertime.preTask to run a selenium task to before the page is analyzed. Documentation is coming soon.
-
New HAR files rock! In the old version we use BrowserMobProxy as a proxy in front of the browser to collect the HAR. In the new version we collect the HAR directly from the browser. For Firefox we use the HAR export trigger and in Chrome we generates it from the performance log.
-
Stability: We have a new completely rewritten version of Browsertime that makes it easier for us to catch errors from the browser, drivers and environment problems.
-
Speed: Yep we dropped Java (it was needed for BrowserMobProxy) and most things are happening in parallel with the new version.
-
Don't overload Graphite: One thing that was annoying with 3.x was that it by default sent a massive amount of metrics to Graphite. That's cool in a way but it was too much. We now send curated metrics by default and you can choose to send more.
-
You can collect metrics from Chrome on an Android phone. In the current version you need to have it connected using USB to the server running sitespeed.io, lets see how we can make it better in the future.
-
Using our Docker container you will get support getting SpeedIndex and startRender using VisualMetrics. This is highly experimental at this stage.
- Dependency problem for PhantomJS 2.1. We hope it works now :)
-
Typos #804 Thanks @beenanner
-
Making slimerjs more indepenedent from phantomjs #806 thanks @keithamus
- Upgraded default PhantomJS to 2.1. We will remove PhantomJS 2.0 from Docker containers.
- Fixes bug for collecting summary metrics for WPT, introduced in 3.11.2
- Browser name in WebPageTest can have spaces and that wasn't handled so when the metrics is sent to Graphite, it fails. #798
- Bumped 3rd party dependencies: winston, request, phantomjs, moment, fs-extra, browsertime, cross-spawn-async, async
- Upgraded Browsertime to new version to work with Node 4
- Report the sitespeed version as full integeres to Graphite. Meaning 3.10.0 will be 3100.
- Normalize file names, use _ as separator in the domain name instead of dots #742
- Do not report skipped rules as failed on the Budget page. thanks @jzoldak #753
- Grunt-sitespeedio fails the build no matter if GPSI score matches the budget thanks @laer #746
- Fixed the ability to supress domain data beeing sent to Graphite. Using --graphiteData you now need to explicit use domains if you want to send the data (if you don't use all). That data shouldn't always be sent as it was before. (thanks @xo4n for pointing that out) #755
- Add ability to budget on a per rule basis, thanks @jzoldak #751
- Add waitScript logic to screenshots thanks agaib @jzoldak #737
- Sorting by url for pages/assets on result pages now works as expected. #743
- Be able to disable the proxy when running BrowserTime #735
- All args to the headless script should be passed in the right order #727 thanks @jzoldak for the PR
- If a site uses SPDY, the sizes in the HAR from WebPageTest is set to 0, don't use it to populate the sizes #699
- Create same Graphite namespace structure for the domain url as for per-page metrics #728 thanks @JeroenVdb for the PR. Use the flag --graphiteUseNewDomainKeyStructure to turn on the new standard (will become default in 4.0).
- You can choose to have query parameters in the Graphite key for the URL, thanks @jeremy-green for the PR #719. To use query parameters in the keys, add --graphiteUseQueryParameters to your run.
- Parameters passed in the wrong order for basic auth when taking screenshots, thanks @jzoldak for the PR! #691
- Use same time out time when taking screenshots and when running yslow #725
- Removed faulty error logging from WPT if your location missed browser configuration. That was wrong, you actually don't need it.
- Basic Auth was missing when testing one page (since 3.7.0). Thank you Jesse Zoldak (@jzoldak) for the PR! #717
- You can now pass configuration files again. When you do a run, the config.json will be in your result folder. Pass it again with --configFile to your next run and it will be tested again, but in a new date result dir #270
- Pass request headers as JSON as complement to all headers in a file #715, thank you Devrim Tufan (@tufandevrim) for the PR.
- Testing a page multiple times ended up with HAR with many many requests for each page. This is now fixed in BrowserTime v0.10.2 #707
- Sending version numbers of sitespeed.io, firefox and chrome removes all dots except the first one
- Show and send total summary of collected data for all pages. Example: testing ten pages we will now have the total number of requsts made for all those 10 pages #693
- Send all individual navigation timings to Graphite (before we only sent calculated timings) #580
- Send sitespeed and browser version to Graphite. The keys in Graphite: [namespace].meta.chrome.version, [namespace].meta.firefox.version and [namespace].meta.sitespeed.version #703
- Bumping versions: async, cross-spawn-async, fast-stats, fs-extra, handlebars, html-minifier, moment, phantomjs, request, winston, browsertime & xmlbuilder.
- If we just test one page, then skip the crawling (but do one request that checks the status of the page before we start) #706
- Text fixes, thanks @atdt #690
- New Browsertime version fixes Browser name and browser version in the HAR file #704
-
Finally we have a HAR Viewer! It's a modified version of Rafael Cesars https://www.npmjs.com/package/simplehar. It could still need some love and work but we think it will add some real value.
-
Bug fix: Worst Cached assets on the hotlist looked bad on small screens.
- Bug fix: Installation works on Windows again. Updated Browsertime with new Selenium that fixes the original problem.
- Bug fix: All WPT metrics wasn't sent ok #681
-
Holy cow, we now support WebPageTest scripting ( https://sites.google.com/a/webpagetest.org/docs/using-webpagetest/scripting)! Every occurrence of {{{URL}}} in your script will be replaced with the URL that is actually going to be tested. Feed the script file to sitespeed.io using --wptScript #623
-
When we are at it: also support custom scripts to collect metrics for WebPageTest! Feed your custom javascript metrics file using --wptCustomMetrics. Read more here: https://sites.google.com/a/webpagetest.org/docs/using-webpagetest/custom-metrics #678
-
Graphite keys now replaces pipe, comma and plus. If your URL:s has them, they will now be replaced by a underscore. The reason is that some Grafana functions don't work with the special characters. Thanks @EikeDawid for the PR! #679
- Rewrite of the Graphite key generation, now always follow the patter protocol.my_domain_com._then_the_path #651. The cool thing is that this opens for a lot og Graphite/Grafana goodies, the bad is that if you use it today, you need to remap the keys in Grafana to get the graphs.
- Super ugly quick fix for dubplicate dots in graphite keys #654
- IMPORTANT: New structure for URL paths sent to Graphite. Now follow protocol.hostname.pathname structure, thanks @JeroenVdb #651
- Send size and type of every asset to Graphite #650 thanks @JeroenVdb!
- Renamed requesttimings to requests when choosing which data that should be sent to Graphite
- Hail the new default waitScript! If you are using phantomjs2 we will now wait for the loadEventEnd + aprox 2 seconds before we end a run for YSlow #653
- IMPORTANT: The old graphite key requests (showing number of requests) changed to noRequests.
- Bug fix: Specifying a custom yslow script now works again, fixing #649. Thanks to Jubel Han for reporting.
- Bug fix: Add size & number of requests for all domains
- Send accumulatedTime a.k.a total time for a domain for all assets to Graphite
- Bug fix: Fetch number of assets used per domain only once per HAR file.
- Bug fix: Number of request per domain always showed 1, that's not right!
- Send the total weight per domain and per content type to Graphite #644 thanks @JeroenVdb.
- Changed the key structure for request per domain (to follow the same structure as the rest of the domain keys): NAMESPACE.www.myhost.com.summary.domains.requests.*
- If you start sitespeed with a callback method it will now call the callback(error, result) now supplying the result.
- Silence log when running in test env (log errors and above)
- Bug fix: Removed another limit for number of domains, setting it to 10 domains before #643
- Removed the limit of sending max 100 domains to Graphite #643
- Use --postURL to POST the result of an analyze to a URL
- Use --processJson to rerun all the post tasks on a result, use it to reconfigure what data to show in the HTML output.
- Bug fix: extra check when generating Graphite keys. #642
- Send request timings to Graphite if we use WebPageTest #639
- Possible to configure the URL to a selenium server (making Chrome on Linux more stable or at least less instable)
- Bumped to new Browsertime version with support for starting your own Selenium server. Use --btConfig to configure the server.
- Bug fix: 3.2.4 contain code that shouldn't be released ...
- Bug fix: Put User Timings (from the User Timing API) in the summary
- Bug fix: Put custom javascript value into the summary
- Bug fix: Budget for BrowserTime metrics stopped working since the last restructure of BT data.
- Bug fix: --storeJson (storing all collected data in one JSON file) didn't work.
- Bug fix: PhantomJS 2.0 had stopped working fetching timings #621, thanks Patrick Wieczorek for reporting
- New Browsertime version, putting User Timings back in the statistics and killing hanging Chrome/Chromedrivers on Linux. Older version could hang when running Chrome on Linux.
- Check that URLs are valid when fetched from a file
- Bug fixes: Compressed sizes has been wrong a long time since a bug in PhantomJS. However, if you also fetch data using browsers or WebPageTest, the sizez will now be correctly populated! #54 #577
- New Browsertime 0.9.2 with fix for HTTPS, making requests visible in HAR-files.
- Check that we got an HAR from WebPageTest before we use the data #596
- We have made it easier to test multiple sites. Add multiple sites by pointing out multiple files like --sites mySite1.txt --sites mySite2.txt. This is done becasue it plays much nicer with our docker images. #579
- Default memory decreased to 256, the old 1024 is only needed when fetching really big sites. 256 is good because it is easier to use oob on small boxes. #610
- Make it easier to use customScripts and waitScript in BrowserTime. Custom scripts metrics is now shown in the result pages and sent to Graphite #611
- Upgraded to latest BrowserTime 0.9 with new structure of the data
- Simplified the proxy usage #612 meaning the proxy will start when Browsertime is needed
- Better title and descriptions #605 and removed robots no index
- Minify HTML output #608
- Bug fix: Handle requests with malformed URI:s when sending data to Graphite #609
-
Send all timings per requests to Graphite when collecting data using Browsertime and WebPageTest (turn on with --graphiteData all or --graphiteData requesttimings) #603
-
Flatten the domain name when sending domain timings to Graphite (making it easier to make nice graphs) #601
-
Configure paths to assets for result pages #604
-
Bug fix: If Graphite server is unreachable, callback chain was broken, meaning sitespeed didn't end properly #606
- New Browsertime version, setting timeout for browser drivers.
- Cleaned up the structure for Graphite internally #600
- Send domain timings info to Graphite to spot slow domains
- Show errors in error page, only when we have errors
- Upgrade to handlebars 3.0
- Upgraded Browsertime with new Selenium version, making Firefox 36 work
- Added verbose logging for GPSI
- Bumped Browsertime version, including fix for stopping Browsermobporxy on Windows
- Running only one run for WPT made aggregators failed (once again) #589
- Links in CLI now pointing to new documentation URL:s
- Log Graphite host & port each time the metrics is sent
- Faulty configuration for default WebPageTest location #588
- Enable verbose logging in Browsertime whenever Sitespeed.io runs in verbose mode (--verbose/-v).
- Check that location for WPT always contains location and browser
- Bumped BrowserTime, new version making sure it will not hang when Selenium/Chromedriver has problems.
- Log the time the analyze of the URL(s) took #578
- Improve validation of command line parameters.
- Added perf budget HTML page #576. Running a budget will also create an HTML page of the result. Thanks @stefanjudis for the idea!
- New BrowserTime version (0.8.22)
- Include Node.js version when printing versions at the start of each run.
- Ok, incredible stupid implementation by me for the current perf budget, throwing an error when failing. Now the result array of the tests are returned.
- Added support for having a budget for number of requests, type and size #571
- Bug fix: Since we added slimerjs, we showed headless domContentLoadedTime as a page colum for every tested page.
- Bug fix: Don't automatically add column data if you configure it yourself.
- Bug fix: Headless timings perf budget was broken.
- Changed to eslint from jshint.
- Updated to latest phantomjs package.
- Updated to latest BrowserTime (with 2.0.0 of BrowserMobProxy)
- You can now choose not to create the domain path in the result dir by using the flag suppressDomainFolder #570
- Better handling of feeding URL:s via file. Supports array with URLs and file.
-
Support for SlimerJS. Note: Choose which "headless" analyzer to use by using --headless [slimerjs|phantomjs]. Default is phantomjs. Also fetch timings using your choice by configure -b headless #544 #559
-
Run WebPageTest test from multiple locations/browsers/connectivity. In 3.0 you could only use one browser/location/connectivity, now you can use as many as you want. Everything is backward compatible except the Graphite keys for WebPageTest has changed, now including browser, location and connectivity.
Meaning you need to change Grafana or what tool you are using to use the new keys when you upgrade. #546
-
Hardcoded dependencies in package.json
- Changed deprecated phantom.args to be compatible to PhantomJS 2 #558
- Bugfix: Errors when taking screenshots weren't recorded as errors.
- Bugfix: Fix crash when running analysis #562
- Choose if you want to create HTML reports or not (--no-html) #548
- Bugfix: URL:s with and without request parameters collided when data files was created, now an extra hash is added to URL:s with parameters #552
- Better logging for PhantomJS
- fixes in the YSlow script so that some pages that fails, will work #549
- Add experimental support for running yslow in SlimerJS #544
- Fix Google PageSpeed Insights that broke in 3.0 #545
- Better logs when screenshot fails and increased timeout to 2 minutes
- Upgraded to new Crawler with higher default timeout times #547
- Added parameter to configure which phantomjs version to use (--phantomjsPath)
- The main goal with 3.0 has been to move to NodeJS. The crawler & BrowserTime still uses Java and we will try to move away from that in the future
- Support for getting Navigation Timing data from PhantomJS 2.0
- Drive/get/collect data from WebPageTest & Google Page Speed Insights
- Two new summary pages: Slowest domains and Toplist (with information about assets). More info will come
- We use Handlebars templates (instead of the old Velocity ones).
- All data is is JSON instead of XML as it was before.
- HAR-files created from the browser you use when fetching Navigation Timing API metrics
- Generate JUnit XML files/TAP and/or send data to Graphite; now included as main functionality.
- Send your metrics to Graphite
- Yep, hate to say it but the parameters to the CLI has changed, so please check --help to see how you should do
- Upgraded to YSlow 3.1.8 (with configurable CDN)
- Added support for Basic Auth (only missing in Browser Time right now, meaning you can't use Basic Auth to get timings)
- Turned on CDN rule and made it possibly to supply CDN domains
- Added new crawler that solves problem when the crawler picked up URLs with wrong domain, thanks @ChrisSoutham
- Updated support for catching lazy loaded assests in YSlow/PhantomJS, thanks @dorajistyle
- Show the 90 percentile value for all timing metrics on the individual pages #380
- Mobile rules changed: Doc size max 14 kb and max server side 200 ms for Green
- Summary: Show max number of request per domain #384
- Summary: Show number of redirects per page #385
- The avoid scaled images rules has been changed: If the image is larger than X (100 pixels by default) the rule will kick in.
- The sitespeed.io-sites script now uses firstPaint as default if you use IE or Chrome, supports local config file & uses maxRequestsPerDomain as default column instead of max requests per domain #387
- DNSLookup hurts more than CSS requests in points for Critical Path Rule #392
- Bug fix: If an error happens when crawling, log that to the error.log #378
- Bug fix: User defined measurements get Velocity code as description #366
- Bug fix: Show one decimal for time metrics on pages #363
- Bug fix: Connect via any SSL protocol #379 thanks @tollmanz
- Kind of a bug fix: The crawler can now handle a href tags with a line break instead of space between the a & the href
- New BrowserTime version 0.6 that fix crash while trying to run resource timing measurements in Firefox, see the list of changes here: https://github.com/tobli/browsertime/releases/tag/browsertime-0.6
- Bug fix: New version of the crawler, the proxy support was broken in the last release
- Added proxy support when collecting Navigation Timing metrics #351
- Added support for local configuration where you can override default configuration (thanks @AD7six)
- Bug fix: If phantomJS fails, the whole analyze fails (introduced in 2.5.x) #359
- The crawler now handles gziped content #263
- When parsing all individual HTML files, show how many that has been parsed every 20 run #354
- Bug fix: The internal link to assets on the detailed page don't work #355
- Bug fix: Redirected URL don't report the end location URL (see the description in the issue for the full story #356)
- Even better fix for #352
- Fixed defect when trying to output error to the current console (instead of using the stderr) #352
- Better error handling: Log all errors to the error log file and #334 and make sure one page error will not break the whole test #329
- Test in multiple browsers in one run #341
- Put the HAR file in the HAR directory instead of sitespeed home dir (fixes #343), now it will work in Jenkins for Ubuntu
- If Chrome or IE is used, display firstPaintTime in the summary as default #307
- Added more default fields in the summary: requestsMissingGzip, jsWeightPerPage, cssWeightPerPage #325
- Changed order of the summary fields so that logical fields are grouped together
- Made it clearer that CSS & JS weight are per file in the summary (meaning inline CSS/JS are not included)
- Show red/yellow/green for cacheTime on the summary page #312 and for JS & CSS size
- Added short description on each rule on the summary page (hover to see it) #161
- New XML-Velocity jar that with a small change how template files are loaded
- New BrowserTime version that works on Windows & fetch resource timings
- Sitespeed.io works (again) on Windows, this time also when fetching Navigation Timing metrics
- Show which browser that is used on the summary page when collecting timing metrics
- Bug fix: the rule "Avoid DNS lookups when a page has few requests" was broken, couldn't tell if JS was loaded async or not #328
- Bug fix: Running the JUnit test script after you fetched URL:s from a file was broken
- Bug fix: Fixed bug When api.exip.org is down (sitespeed stops to work)
- Put the JUnit files into a dir named /junit/ when running the sitespeed.io-junit script. WARNING: this means you need to change in Jenkins where you match the files.
- Include -V when listing supported options in command line help
- Bug fix: The sitespeed-sites script had wrong path to the sitespeed script
- Added the number of text assets that are missing GZIP on the summary page (xml) #310 and for pages #315. Add it by the name requestsWithoutGZipPerPage
- Fix in how to handle browser parameters, to get it to work clean with Jenkins
- Bug fix: User marks named with spaces broke the summary.xml
- Bug fix: Sites with extremely far away last modification time on an asset, could break an analyze
- Upgraded Browser Time version to 0.4, getting back custom user measurements.
- Bug fix: Cleaner handling of relative URL in the sitespeed.io-junit script.
- Moved all scripts to the bin folder, following the standard and easier to package
- Cleanup all scripts to use absolute paths, making it easier to package for Homebrew
- sitespeed-sites.io now always need to have the filename of the text file containing all the URLS
- New names: sitespeed.io-sites & sitespeed.io-junit
- New BrowserTime version (0.3) including backEndTime & frontEndTime
- Changed default summary page to show backend & frontend time (removed redirectionTime & domInteractiveTime)
- Increased timeout for the crawler for really slow pages
- Bug fix: The fix for removing invalid XML caharcters created by GA, sometimes broke the analyze, now fixed (#304)
- New BrowserTime version, having 60s wait for load, also fixes Firefox 25 bug
- Logging all PhantomJS errors to own PhantomJS error log
- Bug fix: URL using brackets didn't get correct doc size
- Bug fix: Unable to crawl websites with GA cookie #298
- Bug fix: sitespeed-sites.io used the sitespeed.io script with sh instead of bash
- Create two JUnit xml files, one for rules & one for timing metrics! The new names of the output files are: sitespeed.io-rules-junit.xml & sitespeed.io-timings-junit.xml
- Finetuned the logo
- Output the the input parameters to the error.log so it is easy to reproduce the error
- Centralized the error logging
- Added an easy way of include sitespeed.io in Travis-CI
- Made it possible to analyze a site with non signed certificates
- Prepared for HTTP 2.0 rules & renamed the current rulesets, new names: sitespeed.io-desktop & sitespeed.io-mobile
- Also copy the result.xml file to the output dir for sitespeed.io-junit.xml (to be able to create graphs per URL)
- Bug fix: The crawler sometimes picked up URL:s linking to other content types than HTML
- Bug fix: The JUnit xslt outputted timings metrics
- You can now choose which data that is showed in the summary boxes (the red/yellow/green ones on the start page) and the columns on the detailed summary page.
- You can also create your own box with your own data on the summary page and your own columns and data on the detailed summary page.
- Fetch Navigation Timing API data from a real browser, using Browser Time.
- New modified rule: the YSlow rule yimgnoscale doesn't work with PhantomJS. The new rule compare the image browser width and the real image width.
- The main script (sitespeed.io) has been cleaned up & Velocity templates has been restructured.
- You can now test multiple sites and compare them using the sitespeed-sites.io script (you can choose what kind of data to compare).
- There are now two different rule-sets, one for desktop & one for mobile.
- You can now configure the limits for the rules on the summary page.
- Phone view on detailed summary page now only contains url & score to make it simpler to maintain.
- You can now see the IP of the computer running the test in the result.xml file.
- You can now set the max pages to test & the name of the test (displayed on every HTML page).
- Simplified user agent by choosing between iphone, ipad or nexus and a real agent & viewport is set.
- Output as CSV: Choose which column to output and always output ip, start url & date.
- Fix for Windows-users that is having spaces in their path to Java.
- Bug fix: URL:s that returns error (4XX-5XX and that sitespeed can't analyze) is now included in the JUnit xml.
- Bug fix: The JUnit script can now output files to a relative path.
- Bug fix: User Agent is now correctly set.
- Supply a test name that will be shown on all pages. Use the the parameter -n
- Well the problem is like this: Today there is no way to get the ttfb from PhantomJS so it is fetched by a extra request using curl. Some sites that don't cache internally (and are slow) can differ quite much for ttfb, meaning ttfb can be higher the next request than the load time the first time. If this happens it is now highlighted.
- Bug fix: show median front/back end time instead of percentile on summary page
- Bug fix: when the ttfb is larger than pageload, don't add it to summary stats
- Bug fix: for some sites (very rarely) the total weight was fetched wrong by YSlow, fixed last release for all pages except summary & summary details.
- Show percentage of requests & size per content type
- You can now export the pages data to csv (again) with the switch "-o csv"
- Upgraded the crawler: Better closing of connections, URL:s that not following RFC 2396 gave null pointers & when a cookie is not following the spec, the url of the page setting the cookie is now logged
- On detailed page summary: Categorize favicon as favicon instead of others, and doc type has now an own category
- If an analysis fails, the url and the error from YSlow is now logged
- Cleanup: When you feed sitespeed.io with a list of urls from a file, the text messages is cleaner in the HTML
- Cleanup: Removed old JS table sorter on pages summary, works better now
- Bug fix: If a page was redirected, the gzipped size was fetched as 0, now fixed
- Bug fix: For some sites (example www.bike.se) the total weight was fetched wrong by YSlow, now the weight is calculated from each assets on the pages info page.
- Bug fix: Urls containing & broke some tests
- Bug fix: In very rare cases, YSlow reports a larger cache weight than the page weight. The bug is not fixed but when it happens, an error message is displayed
- Bug fix: Page weight was sometimes wrongly calculated, now each asset weight is used to calculate
- Bug fix: TTFB on detailed page said ms but the value is s
- Bug fix: Yslow didn't honour max-age before expires (following 1.1 spec), now fixed
- Bug fix: Yslow couldn't fetch headers/weight for @import relative css, fixed
- Bug fix: Yslow now fetch assets, not fetched by PhantomJS
- Added the original favicon rule from Yslow since it works now with the bug fix
- Added favicon to the expire rules
- Changed report dir name. Before it was sitespeed-HOST-NOWDATE, now it is HOST/NOWDATE, to make it easier to compare runs over time
- Added new XML format for the summary page.
- New page: The detailed summary page.
- Always output the result of the sitespeed-junit.io to the data dir
- Changed summary page: Before showed average & median, now median & 95 percentile
- Added summary of total image weight per page & on detailed level you can see individual size
- New rule for checking if old versions of plugins is used. Right now only check JQuery.
- A little better check for correct Java version.
- Bug fix: The check for number of DOM elements where wrong when checking for warning
- Added check that Java exists before the analyze
- Feed sitespeed with either a url to crawl or a plain text file with a list of URL:s (NOTE: the -f argument is now used for the file, the -c is the new for follow a specific path when crawling)
- Create a junit xml file from the test, new script & new xsl file
- Added new max size of a document, using stats from http archive
- Showing the number of domains on page summary
- Showing the percentage of assets that are cacheable on page summary
- Show the amount of assets that don't have an expire header om site summary & pages summary
- Removed prime cache values from site summary & page summary (was not always correct)
- Refactored page summary
- Removed rule ynumreq and created three new ones in order to get clearer junit xml result (and also only check for sync js): cssnumreq, cssimagesnumreq & jsnumreq
- Added average & median nr of DOM elements on site summary and specific nr on page
- Added response headers info on assets page
- Bug fix: If a max age HTTP cache header was missing the cache time, the cache time was set to 0, not listening to Expires header.
- Bug fix: If a asset was missing expire header, the last one before that expire was inherited
- The SPOF rule now only report font face that are loaded from another top level domain. Also the actual font file is reported.
- Show requests per domain on individual page.
- Show cache time for each asset & delta between last modified & delivered + average.
- Configure which yslow file to use and which ruleset.
- Show time spent in frontend vs backend per page.
- On the summary page, now show info blocks: time spent in frontend/backend and cachetime/last modified average.
- New page: The assets page, show the most used assets for the site
- Adjusted the warning rules on the summary page, now a warning is up to the average number collected from httparchive (where applicable)
- Java jar dependencies now compiled for java 1.6 and higher
- Upgraded Yslow to 3.1.5
- Bugfix: The SPOF rule reported CSS & JS as SPOF:s wrongly
- Bugfix: The crawler reported links returning 200 with another content type then text/html as error url:s
- Added support for configuring the crawler (see the dependencies/crawler.properties file).
- Added support for analyze behind proxy (thanks https://github.com/rhulse and https://github.com/samteeeee for reporting and testing it)
- Added html page that shows url:s that returned errors from the crawl
- Added percentage on summary page
- Added support for setting user agent
- Added support for setting view port
- Removed experimental rule for the amount of JS used
- Added new rule: Critical Path
- Finetuned the SPOF rule: Now also check font face
- Added time to first byte (using curl, new requirement)
- Fixed so document weight is fetched from curl aka the right sized if gzipped
- Bugfix: The check for phantomjs wasn't working, works now
- Bugfix: Now using JAVA_HOME in a correct way (thanks Rob-m)
- Bugfix: Upgraded the crawler to 1.3, now only fetched text/html links
- Removed csv as output format
- New rule: Avoid CDN lookups when your page has few requests
- New rule: Do not load stylesheet files when the page has few request
- New rule: Have a reasonable percentage of textual content compared to the rest of the page
- Changed the limit value for doc size on the summary page, vas 10 kb but gzip is taken into consideration, changed to 100 kb!
- Concatenating css & js in the results to one file each
- Show average of how much of a page that consist of javascript in percent, on the summary page
- Show median values where applicable on the summary page (now show both average & median value)
- Show how much of a page that is js & css on a page in percent, compared to content
- Made java heap size & result directory configurable from the sitespeed script
- Cleanup if the sitespeed.io script, removed unused code and made it easier to update
- You can now zip the output result by calling the script
- Upgraded to latest crawler & xml-velocity jar
- Added image, css, js and css image total weight/size on page view
- Added new experimental rule of javascript percentage
- Upgraded jquery from 1.8.2 to 1.8.3
- Made all pages responsive, standard stuff except the table in pages.vm, two different tables, one for phones, one for the rest
- Moved webpagetest link from pages.vm to page, to make more space
- Added the summary data also on page.vm
- Added possibility to output all page data as csv file (yes that old thing)
- Added possibility to output every html file as png, for easy include in documents/web etc
- Added htmlcompressor (http://code.google.com/p/htmlcompressor/) to compress all html, to make the pages smaller
- Upgraded to latest version of the crawler (smaller in size)
- Added support for handling testing only one page (depth = 0), thanks @tomsutton1984
- Better handling of input parameters, now you specify them in the order you like
- Possibility to not crawl specific path segments in urls
- Run multiple processes when analyzing pages (to make it faster)
- More documentation in the sitespeed.io script
- Include rules dictionary when using yslow, always update the doc.js in yslow when adding new rule
- Show the full rule name when showing broken rules
- Show explanation for all rules used, linked from summary page
- Show exact rule number, for easier trackback
- New crawler instead of wget that didn't work on some sites with spider options (amazon etc)
- Fix for css in head rule, now only dns lookups are punished, not the number of css
- Crawl by follow a specific path, meaning you can analyze parts of sites
- Fixed bug that sometimes url:s was fetched from different domains than the main domain
- Added links to tested start url on both summary and page page
- Added parameters to webpagetest run three times by default
- When a SPOF is found, link to webpagetest with SPOF domains activated is used by default
- Show full urls in pages & page to easier understand which url that is analyzed
- Show extra data in modals to make it clearer
- Popover & better texts on summary page
- Cleanup & bug fixes in the bash script, it sometimes failed on some sites when yslow outputted content after the xml
- Added output png:s that can be used on documents
- New rules: Loading js async and finding single point of failure
- Modified expires to skip analytics scripts
- Updated rules texts
- Added new custom rules and modified existing yslow rules.
- Favicon added :)
- Upgraded to jquery 1.8
- Upgraded Twitter Bootstrap to 2.1
- Better title tag on result pages
- Fixed so that long url:s don't break
- Sometimes output xml was broken
- Only fetch content of type html