Change default of --fatal command line arg to terminate on error

[boxydog]

I noticed #3173 is now conflicted and also in the "pelican 5.0" milestone.

So, this is a re-implementation, along with a simple unit test.

I don't know how to test it well, so it's worth comparing carefully to #3173 during review.

Resolves: 3172

• Ensured tests pass and (if applicable) updated functional test output
• Conformed to code style guidelines by running appropriate linting tools
• Added tests for changed code
• Updated documentation for changed code

[boxydog]

Honestly, I don't quite understand why "ignore" is treated differently in one spot (made into ""), and "errors" is passed into a different spot. I was just recreating the functionality in the other PR.

[boxydog]

Ugh, this breaks a lot of unit tests.

[boxydog]

The original PR used "ignore" to ignore, but did not change init to respect that, so also used "". I changed init to respect "" or "ignore".

[justinmayer]

I haven't had time to dig into this topic very deeply, but I believe this change could affect people who rely on the current default. As such, I'm going to mark this as a WIP but only to prevent accidental merge until we are ready, particularly given the presence of a RELEASE.md file here that will trigger the major release automatically.

Before that major release, I would prefer to publish a new release sometime soon with what is currently in main, plus any other non-breaking PRs that make it in.

@avaris: Given your discussion in the previous PR thread, what are your thoughts about the current state of things here?

[egberts]

Yes, this breaks a lot of unit tests, BUT --fatal (and CRITICAL) should be the DEFAULT sys.exit() condition. (on an unrelated note, I too think that ERROR should selectively do sys.exit() like most UNIX programs do, but I will not pursue that now or anytime soon.

There should be no need for --fatal option. However, we can progress toward the sys.exit() piecemeal as needed.

I do believe we should slowly start identifying all those fatals as really "a hill to die on".

[boxydog]

Unfortunately, I don't understand the desired behavior well enough to represent this PR. I just submitted something with some fixes to try to move forward #3173.

Can someone describe next steps here?