wip(msgpack): handle Nil strings

[gigamaax]

I made another change that I'm worried might not be the right direction:

Here
https://github.com/neovim/go-client/pull/184/files#diff-4686cdf50ccf547850a2e89191f11c343519caef6fd7718fd73b7a1480606bd0L3843

I moved away from using the color map based on https://neovim.io/doc/user/syntax.html#ctermbg. I'm not sure if the ctermbg / ctermfg are expected to work with the RGB colors that are in the color map, or if maybe the color map has been enhanced since v0.9? I'm not sure if I'm avoiding the issue with this fix.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.1%. Comparing base (dd77a91) to head (af022d4).
Report is 9 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #184     +/-   ##
=======================================
- Coverage   83.3%   83.1%   -0.2%     
=======================================
  Files         14      14             
  Lines       3132    3134      +2     
=======================================
- Hits        2610    2607      -3     
- Misses       522     527      +5     

Flag	Coverage Δ
linux-1.18-nightly	?
linux-1.18-stable	?
linux-1.19-nightly	?
linux-1.19-stable	?
linux-1.20-nightly	?
linux-1.20-stable	?
linux-1.22-nightly	82.8% <100.0%> (?)
linux-1.22-stable	83.1% <100.0%> (?)
linux-1.23-nightly	82.8% <100.0%> (?)
linux-1.23-stable	83.1% <100.0%> (?)
linux-1.24-nightly	82.8% <100.0%> (?)
linux-1.24-stable	83.1% <100.0%> (?)
macos-1.18-nightly	?
macos-1.18-stable	?
macos-1.19-nightly	?
macos-1.19-stable	?
macos-1.20-nightly	?
macos-1.20-stable	?
macos-1.22-nightly	82.8% <100.0%> (?)
macos-1.22-stable	83.1% <100.0%> (?)
macos-1.23-nightly	82.8% <100.0%> (?)
macos-1.23-stable	83.1% <100.0%> (?)
macos-1.24-nightly	82.8% <100.0%> (?)
macos-1.24-stable	83.1% <100.0%> (?)
windows-1.18-nightly	?
windows-1.18-stable	?
windows-1.19-nightly	?
windows-1.19-stable	?
windows-1.20-nightly	?
windows-1.20-stable	?
windows-1.22-nightly	82.0% <100.0%> (?)
windows-1.22-stable	82.3% <100.0%> (?)
windows-1.23-nightly	82.0% <100.0%> (?)
windows-1.23-stable	82.3% <100.0%> (?)
windows-1.24-nightly	82.0% <100.0%> (?)
windows-1.24-stable	82.3% <100.0%> (?)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[justinmk]

@zchee @glepnir

[brianhuster]

Is this still WIP?

[gigamaax]

I forgot about this. after looking at this again with fresh eyes and a little bit more of a desire to figure it out (rather than just fix) I think I can say that the concern I described in the initial description is not an issue.

for context, I was worried that dropping the colormap wasn't really fixing the failing test but just making it not visible. I now think that the test set up was incorrectly to begin with.

This was the original error:

    --- FAIL: TestAPI/Highlight (0.00s)
        --- FAIL: TestAPI/Highlight/Nvim (0.00s)
            api_test.go:3865: got cterm foreground is 0 but want 16711680
        --- FAIL: TestAPI/Highlight/Batch (0.00s)
            api_test.go:3946: got cterm foreground is 0 but want 16711680
FAIL

the nvim_set_hl api is being called with CtermForegroud and CtermBackground, which take the the cterm colors, not the color map. in order to use the color map, I think we should have been calling with Foreground and Background instead. I haven't gone back to make that change mainly because aside from being able to see "Red" or whatever color, there's not a whole lot of value since we're just checking what was set was actually set.

as for how it ended up as 0, I think nvim_set_hl overflows during the cast to an int16_t -- I didn't see any validation, but I might have just missed it.

anyway, I think this is ready for a review if @zchee or @justinmk would like to take a look. if you don't agree with my analysis, I'm happy to make any changes or look at it some more. some of the tests cold probably get some more love.