Should validate silently overwrite `issues.json`?

[mmarseu]

When the --report-format option is set to anything but stdout, the validate command requires an output path to write the report file. Currently, this path defaults to issues.json, if the --output option isn't given. Leaving aside the fact that this might not be sufficiently clear to the user, what happens if issues.json already exists?

Expected behavior: As a user I would expect the program to abort with an error message.

Actual behavior: The program silently overwrites issues.json.

To be clear, I'm only talking about the case where there is no --output option. If the user provides a filename, it is their responsibility to ensure they don't overwrite anything important. But an implicitly chosen default value shouldn't lead to overwriting an existing file.

I see two possible solutions:

• The simple one: Require the --output option if a report file is needed. This is clear, fixes the problem and places no great burden on the user.
• Alternatively, implement the "expected behavior" described above. This has the least potential to break existing scripts but its a little more work and IMO, choosing an implicit output filename is just not great behavior in any tool.

[italvi]

@mmarseu the reason is that we started with Jenkins being supported and them wanting a JSON with an issues-array, making the name issues.json the most obvious one.

I would also prefer the first solution, as the tool is still in development, i.e. still on 0.y.z, we can be more open regarding breaking changes.