New: clashCompileError

[DigitalBrains1]

Based upon the version proposed by @basile-henry in issue #2020:
#2020 (comment)

Still TODO:

• Write a changelog entry (see changelog/README.md)
• Check copyright notices are up to date in edited files

[martijnbastiaan]

LGTM (assuming copyright / CI fixes). As far as the clash-dev failure goes: it seems to happen sporadically and I can't reproduce it locally. Given that it is not anything user-facing I'd be fine with disabling the check for 9.2.

[DigitalBrains1]

GHC 9.2 complaining about incomplete pattern matches (which actually pointed out a real bug in my code) made me realise it should probably do better on error reporting in the case a user passes some string which cannot be reduced to a literal by Clash. So I did some improvements there. I also trimmed the Changelog entry a bit; I can be a bit of a blabbermouth.

[DigitalBrains1]

Martijn, could you review these changes:
https://github.com/clash-lang/clash-compiler/compare/5fc0f35..9605ab2a43ef392cac0060ebf44ac0e82b606f23

[martijnbastiaan]

I'd like to have the error "foo" instead of undefined, but otherwise LGTM. Feel free to ignore/incorporate the other suggestion.

[DigitalBrains1]

Ah, actually using TermLiteral changes things a lot, I think it's an improvement. Do you agree?

[DigitalBrains1]

Ah, it can be even shorter; that does produce a useless call stack as part of the error but the code is so nice and simple this way.

[edit]
This is the message with the new implementation:

<no location info>: error:
    Clash error call:
    clashCompileError: <your message here>
    CallStack (from HasCallStack):
      error, called at src/Clash/Netlist/Util.hs:876:17 in clash-lib-1.7.0-inplace:Clash.Netlist.Util

Previously the last two lines were not included by virtue of errorWithoutStackTrace
[/edit]

[martijnbastiaan]

Definitely an improvement!