Use hashed file contents as unique test ID

[ozyx]

Issue #1561

Bug

TFS unable to detect the same test failing in a previous build if the path of the *.js file differs.

Since a test's path is used in its FullyQualifiedName, the same test run on two different agents with different drive names (i.e: D:/ vs. C:/ drives) will not be recognized as the same test by TFS.

Fix

Hash the contents of the test file using SHA1 (so as not to slow down test discovery significantly) and use this as a ModuleName / unique ID.

Previous FullyQualifiedId: C:\some\path\to\my\test.js::Suite 1 Test 1::Mocha
New FullyQualifiedId: test[60E439FF4023183BDC8C]::Suite 1 Test 1::Mocha

NOTE: An assumption using this method is that two test files do not exist in the same project with the exact same name and contents.

Testing

1. Create sample test project
2. Copy project to a different path
3. Open both projects
4. The same test in both projects should have the same FullyQualifiedId

This will ensure intended TFS behavior of failed test detection in current vs. previous builds.

Please let me know if there are any improvements or changes I should make. Thanks!

[paulvanbrenk]

Overall this looks pretty good... I'm not sure cutting the hash in half, gives a random enough hash... and I'm sure we'll get flagged for using this algorithm, so we'll have to explain that we're not using it for encryption.

Would GetHashCode be good enough?

[ozyx]

I'm not sure GetHashCode would suffice (I'm assuming you're referencing string.GetHashCode), as the hash needs to be unique to the file contents. Otherwise just using the file path would work fine. Is there an easier / cleaner way to hash the file contents?

[ozyx]

Another way I could think to solve this problem would be:

• Use filepath relative to the solution or project location
  • However, the executing binary doesn't know this info so it seems it would not be so straightforward...

Let me know your thoughts on this. I'm not sure how big of an inconvenience it is getting flagged for crypto, but it seems hashing the file contents is the simplest way to get an ID unique to the test file itself.

[paulvanbrenk]

Yeah, I'm thinking I should just take it as is for 1.3 and see if we can come up with something else for 1.4. I'll take another look this week and either merge it or explain any changes I want.

Thanks!

[ozyx]

Hi guys-- just wondering if you've had a chance to review this yet? We are using a private build internally currently and so far, so good. But would be nice to get an official release.
Thanks for your time!

[paulvanbrenk]

I'll get a build out later today.

[ozyx]

Great, thanks!

[paulvanbrenk]

Here you go, please let me know if you encounter any issues.

/~https://github.com/Microsoft/nodejstools/releases/tag/v1.3.1