doc: correct description of System Error errno property

[pabigot]

Checklist

• tests and code linting passes
• documentation is changed or added
• the commit message follows commit guidelines

Affected core subsystem(s)

doc

Description of change

Fix issue #6649

[evanlucas]

Thanks Peter

/cc @nodejs/documentation

[estliberitas]

Should we not consider errors created via util._errnoException() as System Errors? Have a look, please:
https://github.com/nodejs/node/search?l=javascript&q=_errnoException&type=Code

The potential problem is that util._errnoException() assigns same value to code and errno:
https://github.com/nodejs/node/blob/master/lib/util.js#L950

And if in some cases errors created by this function will be considered as System Errors, users will assume errno is an integer while it may happen to be a string.

LGTM except the question.

[pabigot]

@estliberitas Good point.

From my perspective what util_errnoException() is doing is wrong, probably the origin of the inconsistent documentation, and very likely hard to fix as the integer value is discarded long before it's invoked.

[jasnell]

Also consider that this is likely going to change a bit following #6573. The code property will be expanded to include node's own errors, there will not be a corresponding errno property with those, and the code values will not be prefixed with E. Perhaps we should wait on landing this one until that one is a bit more settled?

[pabigot]

@jasnell I'm fine with waiting. It's clear there's no simple description that covers what Node.js currently does.

I'm less fine with changing the string representation stored in code from the POSIX standard names like EBUSY in cases where the error corresponds to a POSIX error value, but that's something I'll raise in #6573 if it really bugs me.

[jasnell]

#6573 won't change any of the POSIX standard error names, it just introduces new error codes for things that currently don't have them.

[jasnell]

Closing given the lack of forward progress on this