Update install.bat

[tilkinsc]

The problem here is that %errorlevel% is set on exit /b ]] and you pass it something funny. By adding rem=rem you can reuse a hack and get it still to execute correctly. You don't actually need to even set rem=rem you can just pass it nil, since nothing else in the script uses rem.

[codecov-io]

Codecov Report

Merging #958 into master will decrease coverage by 4.65%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #958      +/-   ##
==========================================
- Coverage   83.59%   78.94%   -4.66%     
==========================================
  Files          90       88       -2     
  Lines       10696     8946    -1750     
==========================================
- Hits         8941     7062    -1879     
- Misses       1755     1884     +129

Impacted Files	Coverage Δ
src/luarocks/admin/cmd/remove.lua	29.09% <0%> (-63.9%)	⬇️
src/luarocks/fs/lua.lua	25.13% <0%> (-62.5%)	⬇️
src/luarocks/admin/cmd/add.lua	36.25% <0%> (-56.7%)	⬇️
src/luarocks/admin/cmd/refresh_cache.lua	53.33% <0%> (-41.41%)	⬇️
src/luarocks/cmd/which.lua	62.5% <0%> (-37.5%)	⬇️
src/luarocks/core/sysdetect.lua	41.33% <0%> (-26.12%)	⬇️
src/luarocks/admin/cache.lua	58.49% <0%> (-22.65%)	⬇️
src/luarocks/cmd.lua	75.47% <0%> (-12.88%)	⬇️
src/luarocks/core/path.lua	82.35% <0%> (-11.77%)	⬇️
src/luarocks/cmd/config.lua	80% <0%> (-11.52%)	⬇️
... and 51 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9a200cf...436ea03. Read the comment docs.

[Tieske]

rem=nil alters the Lua environment if rem was already set to something else. The original rem=rem is better imo, since it does not alter the Lua environment.

Also if exit /b ]] returns a weird exit code (I didn't test it), then surely the right solution is to use exit /b and not exit /b 0 because in the latter case it always returns exit code 0, where in the former case it would leave the exit code unaltered and return the exit code that LuaRocks exited with, which is the right thing to do imo.

Or even better, make it explicit and do exit /b %ERRORLEVEL%.

[tilkinsc]

Ahh thats true.
The ]] still gets processed as an argument.

Also, whats up with the dramatic code coverage declination?

[Tieske]

Also, whats up with the dramatic code coverage declination?

no idea, did you use the latest master branch to create this PR?

[tilkinsc]

Yes I did. I edited directly from the website.

[Tieske]

nvm the coverage for now, it needs updating according to previous comments

[tilkinsc]

The changes were made for rem=rem

[Tieske]

But it is still forcing the exit code to 0 instead of returning the result from the call into Lua

[hishamhm]

Do we still need this?

Looks like exit /b (with or without %ERRORLEVEL%) is a big rabbit hole, which I'm not keen on getting into:
https://stackoverflow.com/questions/66248276/difference-between-exit-b-and-exit-b-errorlevel

And from Lua's perspective that second added rem=rem is inside a Lua string, so it should never run, so I don't think that's necessary either.

But in practice, I don't think a lot of people rely on our install.bat's produced errorlevel (especially if it doesn't currently work!), and to be honest I'd rather eventually remove this script in favor of other installation methods.

If the current behavior is producing a spurious error message thanks to exit \b ]] I'd be happy to just change it to plain exit \b and put the ]] on its own line below.

If this is just producing a bogus errorlevel which goes mainly unnoticed, then I don't care much either way and I would just close the PR.