Introduce configuration option to detect gradle versions from gradle-wrapper.properties

[mikepenz]

I do understand that this PR does not match the "contribution guideline" specifically on the requirement for:

It is not the goal of this action to verify that the gradle-wrapper.jar matches a specific version of Gradle, nor that the version declared in the build.gradle or gradle-wrapper.properties file matches.
https://github.com/gradle/wrapper-validation-action/blob/56b90f209b02bf6d1deae490e9ef18b21a389cd4/CONTRIBUTING.md#non-goals

However, seeing related issues still being kept open - and overall comments indicate positive interest towards such a functionality (E.g. #96 (comment), gradle/actions#286)

It would be great to see this functionality being added as I believe this will speed up the validation as a whole, but also lower the network utilization. Especially as recently the occurrences for request timeouts seem to increase (with builds frequently failing) - Related tickets # 46, # 40,

---

This PR introduces a new configuration option (which is disabled by default) called detect-version.

When enabled the action will now try to retrieve the gradle version associated with a gradle wrapper via the distributionUrl in the gradle-wrapper.properties.

Furthermore, it adjusts the checksum fetching condition to only retrieve checksums for the detected versions (instead of fetching versions for all possible releases).

This PR aims to fix to open issue: #96

However, while it was not specifically designed for this purpose it resolves partially gradle/actions#282 as it will only include checksums for versions detected.
A future update could include that we retain wrapper <-> version && checksum <-> version details to only validate the specific combination.

Additionally, I believe that the initial discovery and parsing of the gradle-wrapper.properties unlock and enable future expansions to for example handle: gradle/actions#286

---

The PR includes the relevant changes to the action and updated test cases to cover the new functionality.

[Marcono1234]

it partially resolves gradle/actions#282

GitHub detected the "resolves #..." as keyword and linked the issue, so it would be closed if this PR here is merged. I assume this is not intended since as you said it only partially resolves it.
This can possibly be solved by simply rewording it, for example from "partially resolves #..." to "resolves partially #...".

[bigdaz]

@mikepenz apologies for the extended delay responding to this. I'm notionally responsible for maintenance of this action, but haven't had time to make it a priority.

[bigdaz]

The changes to dist/index.js make the PR pretty much impossible to merge. Can you please update the PR, removing this change? It's easy for me to regenerate the outputs just prior to merging.

[mikepenz]

@bigdaz no worries. Thank you for having a look at it.
I rebased the branch on top of the latest main and removed the pre-compiled dist/index.js file.

[bigdaz]

@mikepenz @Marcono1234 @JLLeitschuh With the change to include hardcoded checksums, do y'all feel that this is still desirable? I haven't had time to review these changes.

It seems that the upgrade to Node 20 may have made the network situation worse, so my current thinking is that we promptly release v2.1.0 with only the hardcoded checksums fix, and consider this PR in the following release.

Further, has this PR been updated to take advantage of the hardcoded checksums?

[JLLeitschuh]

It seems that the upgrade to Node 20 may have made the network situation worse, so my current thinking is that we promptly release v2.1.0 with only the hardcoded checksums fix, and consider this PR in the following release.

Oof, well in that case, absolutely cut a release ASAP.

Happy to consider this PR later

[mikepenz]

I rebased this PR before the hardcoded checksum PR was merged. Need to check again as it appears to be in significant conflicts now.

[bigdaz]

Oof, well in that case, absolutely cut a release ASAP.

Yep, see these comments from @deejay1: #174 (comment)

[Marcono1234]

@mikepenz, the new KNOWN_VALID_CHECKSUMS is a Map<string, Set<string>>, mapping from checksum to a set of version names. This should hopefully allow integration with your changes here. You could then for example calculate the checksum for the JAR, and check whether the expected version is in the Set for that checksum.
Though if the Map does not contain an entry for it, or if the Set does not contain the version, you would still need to fetch the checksum from the API since new Gradle versions might have the same wrapper checksum as previous versions.

[JLLeitschuh]

When enabled the action will now try to retrieve the gradle version associated with a gradle wrapper via the distributionUrl in the gradle-wrapper.properties.

I'm concerned about relying upon this generally.

In my experience, many people don't know how to use the wrapper task properly. In order to actually update the gradle-wrapper.jar file, you actually need to run the wrapper task twice, the first time to update the gradle-wrapper.properties, and the second time to update the gradle-wrapper.jar. Many don't so the version in the gradle-wrapper.properties is actually one version ahead of whatever version of the gradle-wrapper.jar is being used.

Given this, I'm concerned that this change will significantly increase the false-positive rate of this action, which will decrease user trust in it.

If the hard-coded checksums solve the problem, I'd be more inclined to see if that solves most of the pain before we attempt including this strategy as well.

If we were to move forward with this strategy, I'd advise that we instead use the distributionUrl as a starting point, and then perhaps do a search starting with the distributionUrl version as a starting point.

[mikepenz]

Thank you @JLLeitschuh for looking into this PR.

I rebased it just in case, however feel free to close it, given the hardcoded checksum solution may solve it now.

It may be valid to re-work this PR to offer a "strict" mode which would be more restrictive than just accepting any hash from the whole set. With the intention to force users to have the correct hash for the correct version. With some more messaging to inform that the double wrapper update may be the cause of the issue.

[nooboobhuh]

UJI