Initial implementation of parseCrashes()

[arturocuya]

The base tool structures only the paths found in the crash report. The goal of this PR is to also structure the remaining information for each crash:

• The error message
• The hardware platforms where the crash occured and their count. Not only the code name but also additional information (taken from here) like the product name and the model, e.g: Roku Ultra, model 4800X.
• The application versions and the error count for each.
• The stacktrace, with the scope names and locations of each scope level
• The local variable names and their metadata

This structured list of crashes is stored in CrashlogFile.crashes. Now CrashlogFile.crashes and CrashlogFile.references have redundant information, but I have not removed the latter because I don't know much about the codebase outside of CrashlogFile.

I have left some TODOs with questions. Please address them adding comments to the PR.

[arturocuya]

@TwitchBronBron What about the redundant information in CrashlogFile.crashes and CrashlogFile.references?

[TwitchBronBron]

@TwitchBronBron What about the redundant information in CrashlogFile.crashes and CrashlogFile.references?

There might be situations where there are pkg paths not found inside a crash. The references collection handles every pkgPath in the entire crashlog.

Perhaps for CrashlogFile.crashes, you could add a property called location that points to the specific CrashlogFile.references entry. So instead of crash.pkgPath, you would do crash.location.pkgPath. That way, you get the sourcemap lookups for free since that location would be a reference to the existing CrashlogFile.references item.

[arturocuya]

@TwitchBronBron Update:

• Improved stack trace parsing and added tests for it.
• Removed unnecesary loop.
• Turned off linter rule @typescript-eslint/no-loop-func
• Renamed StackTraceStep to StackFrame
• Replaced StackFrame's pkgLocation and srcLocation for reference which is a pointer to a CrashlogFile.references entry.

[arturocuya]

@TwitchBronBron I honestly couldn't think of a more elegant solution for this. I wanted to make the subparsers for each crash section agnostic to the order of the sections, so Roku can miss sending, for example, the hardware information and the parser could still figure out the stack trace. This resulted in the subparser for the stack trace not having the original line number for each of its section lines. Hence this code...

[TwitchBronBron]

If it's accurate, then I'm fine with this for now. We can improve this over time.

[TwitchBronBron]

Everything looks good here. Is there anything else to do before I merge this? The only thing I can see is to get the CI failures fixed.

[arturocuya]

Everything looks good here. Is there anything else to do before I merge this? The only thing I can see is to get the CI failures fixed.

@TwitchBronBron Yep, just wait for me to fix the coverage problem.