Serious regression from PR 46

[alerque]

Paging @arichard4, we have a small problem. Everything seemed pretty good but I'm doing some pre-release testing and throw the HEAD version at a larger project (a couple hundred Lua files) and end up with something like this:

$ luacheck .
Critical error: Luacheck 0.25.0 bug (please report at https://github.com/lunarmodules/luacheck/issues):
/usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:73: attempt to index a nil value (upvalue 'chstate')
stack traceback:
        /usr/share/lua/5.4/luacheck/utils.lua:187: in metamethod 'index'
        /usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:73: in upvalue 'handle_nodes'
        /usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:69: in upvalue 'handle_nodes'
        /usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:69: in upvalue 'handle_nodes'
        /usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:69: in upvalue 'handle_nodes'
        /usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:69: in upvalue 'handle_nodes'
        /usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:43: in upvalue 'handle_nodes'
        /usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:57: in upvalue 'handle_nodes'
        /usr/share/lua/5.4/luacheck/stages/unwrap_parens.lua:96: in field 'run'
        /usr/share/lua/5.4/luacheck/stages/init.lua:71: in function </usr/share/lua/5.4/luacheck/stages/init.lua:69>
        (...tail calls...)
        [C]: in function 'xpcall'
        /usr/share/lua/5.4/luacheck/utils.lua:201: in field 'try'
        /usr/share/lua/5.4/luacheck/check.lua:49: in function </usr/share/lua/5.4/luacheck/check.lua:47>
        (...tail calls...)
        /usr/share/lua/5.4/luacheck/multithreading.lua:56: in function </usr/share/lua/5.4/luacheck/multithreading.lua:45>
        (...tail calls...)
        [C]: in function 'xpcall'
        /usr/share/lua/5.4/luacheck/utils.lua:201: in field 'try'
        /usr/share/lua/5.4/luacheck/multithreading.lua:61: in function </usr/share/lua/5.4/luacheck/multithreading.lua:60>
stack traceback:
        /usr/share/lua/5.4/luacheck/utils.lua:187: in function </usr/share/lua/5.4/luacheck/utils.lua:183>
        [C]: in function 'error'
        /usr/share/lua/5.4/luacheck/check.lua:62: in function </usr/share/lua/5.4/luacheck/check.lua:47>
        (...tail calls...)
        /usr/share/lua/5.4/luacheck/multithreading.lua:56: in function </usr/share/lua/5.4/luacheck/multithreading.lua:45>
        (...tail calls...)
        [C]: in function 'xpcall'
        /usr/share/lua/5.4/luacheck/utils.lua:201: in field 'try'
        /usr/share/lua/5.4/luacheck/multithreading.lua:61: in function </usr/share/lua/5.4/luacheck/multithreading.lua:60>
stack traceback:
        /usr/share/lua/5.4/luacheck/utils.lua:187: in function </usr/share/lua/5.4/luacheck/utils.lua:183>
        [C]: in function 'error'
        /usr/share/lua/5.4/luacheck/multithreading.lua:98: in function 'luacheck.multithreading.pmap'
        /usr/share/lua/5.4/luacheck/runner.lua:224: in method '_add_new_reports'
        /usr/share/lua/5.4/luacheck/runner.lua:262: in method '_get_reports'
        /usr/share/lua/5.4/luacheck/runner.lua:333: in method 'check'
        /usr/share/lua/5.4/luacheck/main.lua:308: in function </usr/share/lua/5.4/luacheck/main.lua:260>
        (...tail calls...)
        [C]: in function 'xpcall'
        /usr/share/lua/5.4/luacheck/utils.lua:201: in function 'luacheck.utils.try'
        /usr/share/lua/5.4/luacheck/main.lua:343: in main chunk
        [C]: in global 'require'
        /usr/lib/luarocks/rocks-5.4/luacheck/dev-1/bin/luacheck:2: in main chunk
        [C]: in ?

This is using 0.25.0.r42.gca770a4, I don't have the same problem on the same project on 0.25.0 built using the same version of Lua.

I haven't done much to track this down yet, but it is going to need fixing before I can tag 0.26.0 and if you have any ideas I'm all ears.

[alerque]

Further info: every file in the project will pass just fine if run one by one, the issue is with processing multiple files. Either luacheck . or passing a list of 5 or so files will cause this. Some sets of 5 pass some don't. I suspect this is a race condition with lanes, but I don't know that — I just thought I'd through out what I have so far.

[alerque]

Yes, luacheck -j1 . works on everything I can throw at it, -j2 starts dying.