This project is a modified for public version of chewbotkah which runs front-end Quality Assurance (QA) testing on https://www.s2disk.com/. chewbotkah is a little more involved at s2disk.com but I hope that it will give you a decent place to start for your own front end testing. Automate all the things!
Install Ubuntu 13.10 (Server or Desktop) in a VirtualBox Virtual Machine.
First run setup.sh as root. Then run setup.sh as a normal user.
sudo ./setup/setup_firefox.sh
./setup/setup_firefox.sh
#run some basic tests against example.com
./frontend_test.py
./setup/setup_firefox.sh kill
If you run setup.sh twice it should autodetect running processes.
It crawls a frontend and attempts to run basic quality assurance tests in stages.
Currently there are only three stages.
- Crawls a domain using a domain filter so it doesn't attempt to crawl the whole internet. It will grab every unique URL it can scrape. This builds an index of domain specific pages matching the
--domain-filteroption. Each index contains a set of links found on the page. - Run individual test suites on the crawled data.
- Triage the results so they can be resolved in a sane manner.
Each stage and steps in between are optional for the most part. If you have crawl data saved you can skip crawling. You can also skip individual unit tests. The triage report is not done by default.
The testing stages are organized into Suites. Currently there are only 3 suites.
- Suite 1 - This suite analyzes crawler data and checks for links that are not approved through whitelist or do not match the
--domain-filter. This operates on thehrefattribute of<a>element in the scrape data. This does not actually perform any network requests but uses the crawl data. - Suite 2 - This suite loads every hyperlink reference on every page and checks for bad links in the HTML. This test may go off site to test a link.
- Suite 3 - This suite profiles the loading of each page on the domain from crawler data and determins if there are any non-200 HTTP status resources loading on each page. This only tests the crawler indexes and does not check links within pages.
The following options are available for frontend_test.py.
-t URL,--target-url=URL- This is the target page in which the crawler will start.-d STRING,--domain-filter=STRING- This filter stops the crawler from traversing the whole web. This will restrict the crawler to a url pattern.-w LIST,--href-whitelist=LIST- This is a comma separated list which enables a whitelist of any href links that don't match the--domain-filterto pass and all other references to fail. Part of test Suite 1.--request-delay=SECONDS- Delay all requests by number of seconds. This number can be a floating point for sub-second precision.--skip-suites=NUM- Skip test suites to avoid running them. Comma separated list of numbers or ranges.--save-crawl=FILE- Save your crawl data to a JSON formatted file.--load-crawl=FILE- Load JSON formatted crawl data instead of crawling.--crawler-excludes=LIST- Comma separated word list. If word is in URL then the crawler won't attempt to crawl it.--triage-report=FILE- Generate a report with all errors triaged in a markdown format.
ERROR: There was an unexpected Alert! [An error occurred! TypeError: a is null]- This likely means that DOM elements were changed or not ready. While all requests wait to execute until after DOMReady they can't account for JavaScript manipulating the page. To mitigate this use the--request-delayoption.
Because this software uses a modified component of selenium-profiler it must be licensed GNU GPLv3 if it is redistributed. This is because selenium-profiler is licensed under GNU GPLv3.