Do not catch and ignore fatal exceptions

[pjfanning]

I was pitching for this in #3826

I don't think it is a good idea to risk messing with fatal errors like OutOfMemoryErrors.

If this util seems ok, it could in theory be moved to jackson-core and used in other Jackson modules.

fyi @JooHyukKim

[cowtowncoder]

I am not sure LinkageError should be included in this list: isn't it one of failures that specifically can occur on various platforms trying to load optional extensions?

[pjfanning]

I wouldn't be against removing LinkageError here or being specific. Some LinkageErrors seem more fatal than others - see the subclasses here - https://docs.oracle.com/javase/7/docs/api/java/lang/LinkageError.html

NoClassDefError seems least fatal but in use cases where I added this util to be used, we tend to have the Class already.

[cowtowncoder]

Yeah the concern I have is specifically that the optional loading should not die on missing classes; that's pretty much the main use case. I wish actual test case. But conversely I haven't seen reports where catching one of fatal errors here have been problematic -- theoretically I could see InterruptedException catching being problematic, of these I think.

[cowtowncoder]

Ok. I guess that's fine, except for my concern wrt LinkageError.

But how about adding convenience method of rethrowIfFatal(t) to reduce boilerplate code on caller?

[JooHyukKim]

But how about adding convenience method of rethrowIfFatal(t) to reduce boilerplate code on caller?

I was about to suggest the same thing 😂. +1 from me. I like the logic inside `isFatal. It seems safer than other way around.

[pjfanning]

Ok. I guess that's fine, except for my concern wrt LinkageError.

But how about adding convenience method of rethrowIfFatal(t) to reduce boilerplate code on caller?

I've refactored to use rethrowIfFatal(t).

I still haven't changed LinkageError.

[pjfanning]

I've changed the 'fatal' check to ignore NoClassDefError

[cowtowncoder]

Ok looks good. Will merge.