`internal error: entered unreachable code` when an import path is a symlink

[twitchyliquid64]

Describe the bug
I switched over a directory which was a full directory to be a symlink to something 2 components down + across, and was greeted with all my tests failing with:

internal error: entered unreachable code: record type was Var(LocIdent { ident: `spec`, pos: Original(RawSpan { src_id: FileId(3), start: ByteIndex(798), end: ByteIndex(802) }), generated: false })

Its hard to trace with the whole internal error wrapping, but my best guess is that Cache::normalize_rel_path or Cache::normalize_abs_path is hitting the match arm for Prefix.

Anyways, ill switch back to using a directory for now.

[yannham]

Thanks for the report. What version of nickel are you running? Does the unreachable provides a source file and a line number in the panic message?

[twitchyliquid64]

We are running 0.15.1! And unfortunately there was no such detail - it seemed it go swallowed.

[yannham]

It's really strange; I don't get where an unreachable! panic is transformed into an internal error. Can't find the message record type was anywhere either. Are those coming from your own layer that uses nickel_lang_core in some way?

[yannham]

What platform are you on?

[twitchyliquid64]

Debian Linux, old stable

[twitchyliquid64]

It's really strange; I don't get where an unreachable! panic is transformed into an internal error. Can't find the message record type was anywhere either. Are those coming from your own layer that uses nickel_lang_core in some way?

This is what puzzled me too - I'll take a pass through my code to make sure that its not unreachable! in my code that was firing.