bisect-builds.py
Have you ever hit a regression bug like this: "In chromium 85.0.4183.121, things were broken. Back in chromium 86.0.4240.193, it was fine."? A good way to attack bugs like this – where it's unclear what change could have caused the regression, but where you have a reliable repro – is to bisect.
tools/bisect-builds.py automates downloading builds of Chrome across a regression range, conducting a binary search for the problematic change.
If you don't have a chromium checkout, you can fetch just this script with the commands below:
Linux
curl -s --basic -n "https://chromium.googlesource.com/chromium/src/+/HEAD/tools/bisect-builds.py?format=TEXT" | base64 -d > bisect-builds.py
macOS
curl -s --basic -n "https://chromium.googlesource.com/chromium/src/+/HEAD/tools/bisect-builds.py?format=TEXT" | base64 -D > bisect-builds.py
Windows (no curl, base64) python3
python3 -c "import base64; import urllib.request; print(str(base64.b64decode(urllib.request.urlopen(\"https://chromium.googlesource.com/chromium/src/+/HEAD/tools/bisect-builds.py?format=TEXT\").read()),'utf-8'))" > bisect-builds.py
Run it like this:
python tools/bisect-builds.py -a platform -g good-revision -b bad-revision -- flags-for-chrome
For example,
python tools/bisect-builds.py -a mac -g 782793 -b 800218 --use-local-cache --verify-range -- --no-first-run --user-data-dir=/tmp http://example.com
Also, you can specify either end of the bisect range using version numbers.
python tools/bisect-builds.py -a mac -g 85.0.4183.121 -b 86.0.4240.193 --use-local-cache --verify-range
The two range specifications above are equivalent. Note that in all cases the bisect is being done from trunk builds so merges to a release branch will not be bisected.
Note: bisect-builds.py script now requires Python 3 to run for Windows. Use python3 on Windows instead, in order to run the examples above.
Valid archive types (the -a parameter) are mac, mac64, win, win64, linux (not supported for builds after March 2016), linux64, linux-arm, and chromeos
.
You can also use the -p
option to specify a profile. If no -p
or
--user-data-dir
option is specified, a new profile will be created in a
temporary directory each time you are asked to try a build. If you specify a
profile folder, point to the directory that's a parent of Default/.
The script will download a build in the revision range and execute it. You must then manually check if the bug still repros. Quit Chromium, and the script will ask you if the bug reproduced or not. It will use your answer to drive a binary search, and after just a few steps it will tell you "this regression happened somewhere between revisions 793241 and 793248". From that list, it's usually easy to spot the offending CL. If you're adding the range as a comment to a bug, please always paste the output from bisect-builds.py, as this includes links to the chromium changes in the regression range.
View code changes in revision range with this useful URL (replacing SUCCESS_REV and FAILURE_REV with the range start and end):
https://test-results.appspot.com/revision_range?start=SUCCESS_REV&end=FAILURE_REV
Notes: The default option is snapshot(Chromium build). There are also release build(-r) and official build(-o) available for Googlers. Googlers should use official build(-o) whenever possible to bisect to a single commit. Please refer to go/chrome-bisect for more information.
Getting an initial revision range
If you have two Chrome binaries, one which doesn't work, one which does, you can find their revision numbers as follows.
First, visit the [chrome://version](javascript:void(0);) page and copy the version number (for example 85.0.4183.121). Then, use this version number as the parameter to -g or -b. Alternately you can use stable-release milestone numbers (M85, M86) or look in git commit messages for "Cr-Commit-Position: refs/heads/main@{%825204}".
Verifying the range
If your revision range is incorrect, or if something about your environment interferes with your reproduction of the bug, you will not get useful results from bisect-builds.py. If you would prefer to know this as soon as possible, rather than after downloading and checking O(log n) builds, pass the --verify-range option to bisect-builds.py. This will check the first and last builds in the range before starting the bisect.
Bisecting Browser Automation Tests
You can also use this script for tests built with browser automation frameworks like Puppeteer and Selenium to find the regression range that causes the failure.
First, you'll need to alter your test launch settings to accept
executable_path
(the actual configuration might differ; please refer to the
documentation for your browser automation framework) for the Chrome binary from
environment variables or command-line parameters.
Then, you can run your automated tests and pass the Chrome binary path with the
%p
placeholder in --command
. You might also need the corresponding WebDriver
for some frameworks. Pass --chromedriver
to download it and use %d
in
--command
. Additionally, use --no-interactive
to automatically detect test
failures and enable -v/--verbose
to show the logs from your tests. Here's an
example of setting the ChromeDriver path through an environment variable and the
Chrome binary path via the command line:
python3 tools/bisect-builds.py -g 85.0.4183.121 -b 86.0.4240.193 -v --verify-range --not-interactive -c "CHROMEDRIVER_EXECUTABLE_PATH=%d npx test --browser %p"
Field trials
The bisect script uses non-Chrome branded builds and therefore uses the field
trial testing config.
This means that behavior may differ from what you see in Chrome. Consider
using the disable-field-trial-config
command line switch if this matters to
you.
To find out which variation causes a regression, you can use bisect-variations.py
API Keys and Chrome OS builds
Without API keys, Chrome OS won't allow you to log in as a specific user. To run a chromeos bisect on your Linux desktop, add the following variables to your environment (e.g., via .bashrc):
GOOGLE_API_KEY=<key>
GOOGLE_DEFAULT_CLIENT_ID=<id>
GOOGLE_DEFAULT_CLIENT_SECRET=<secret>
See https://www.chromium.org/developers/how-tos/api-keys for more info about API keys.