Fix Nullpointer for Gradle Parsing

[sconvent]

This will fix issue #6384

The problem was that Dependabot initially recognized correctly that there is no build.gradle.kts in the root folder, but later tried to access it anyways (See top_level_build_file and the comment above here)

This problem is solved with an additional .compact() statement, which will remove any nil entries in the list of files to be looked into.

Tested it by executing it on (an old state) of one of my repos with:
bin/dry-run.rb gradle sconvent/dependabot-issue-demonstrator --commit addf67d1aa51dade24e47d1043abbd0baa54332a

Works as expected (i.e. no Nullpointers)

It will probably also fix #4793, as it is a very similar issue.

Happy to get some feedback as this is my first PR for Dependabot.

[jeffwidman]

LGTM, however could you add a unit test for this scenario to prevent future regressions?

Odds of regressing on this exact problem are low, but the fact that we're hitting an error here shows we don't have much (any?) test coverage for this scenario of a missing top-level file... and odds of regressing in some other way for this scenario are high enough that we ought to get that test in place.

It should be straightforward to write by copying one of the existing tests (maybe one of the ones added in #4256?), but if you run into questions / need help please let me know.

[jeffwidman]

I checked, and FileParser.find_includes() already protects against hitting NullPointer:

dependabot-core/gradle/lib/dependabot/gradle/file_parser.rb

Line 51 in 5dc907e

return [] unless buildfile

[sconvent]

Linting issue should be fixed now.

Sure, I'll add a testcase. Thanks for the hint where I can find one.

[jeffwidman]

Thanks. Feel free to squash your commits unless they're logically separate, or I'm also happy to do that at merge time. We aren't sticklers for single commit PR's esp if multiple discrete changes need to land in a single PR, but no need for a distinct commit just to fix a lint issue.

[sconvent]

Added a testcase (for each Groovy and Kotln DSL) and squashed everything into one commit.
The testcases are a bit different from the one you mentioned. @jeffwidman . But they failed without the changes in the actual code and now succeed.

[jeffwidman]

Thank you! Tests look reasonable to me.