Busted and LuaJIT are not playing well together.

[jszakmeister]

I just wanted to give you guys a heads up that some of Busted's more recent changes have caused some issues for Neovim's test suite using LuaJIT.

This fixes one issue, but I don't think it's the right fix--I think LuaJIT is broken here:

diff --git a/busted/compatibility.lua b/busted/compatibility.lua
index 7837656..3ae58e5 100644
--- a/busted/compatibility.lua
+++ b/busted/compatibility.lua
@@ -55,6 +55,6 @@ return {
         end
       end
     end
-    os.exit(code, close)
+    os.exit(code)
   end,
 }

Some other newer behavior in Busted is now helping to trigger an assertion error about ctypes in LuaJIT too--again, I don't think it's technically Busted's problem, but it may affect anyone who is using LuaJIT and the ffi module.

More than anything, I wanted to let you know that there are some issues with Busted and LuaJIT. I've brought this to the attention of Mike Pall as well. Hopefully, he can help track some of this down too: http://www.freelists.org/post/luajit/Assertion-error

[o-lim]

According to the change log, os.exit(status[, close]) should be supported since LuaJIT 2.0.0-beta6. Sounds like your suspicions may be right on, at least regarding os.exit.

LuaJIT 2.0.0-beta6 — 2011-02-11

• New features:
• PowerPC/e500v2 port of the LuaJIT interpreter is complete.
• Various minor features from Lua 5.2: Hex escapes in literals, '*' escape, reversible string.format("%q",s), "%g" pattern, table.sort checks callbacks, os.exit(status|true|false[,close]).
• Lua 5.2 __pairs and __ipairs metamethods (disabled by default).
• Initial release of the FFI library.

[o-lim]

@jszakmeister: I looked at one of your test cases here (https://www.freelists.org/post/luajit/Assertion-error,2). You are doing a require 'ffi in your spec files, which seems to be related to your error.

If I comment either one of the require('ffi') lines out, it runs
fine. Perhaps busted is clearing something, and ffi is actually
getting loaded twice, even though it shouldn't?

I should add that this happens with busted 2.0-rc7. If I use
busted 2.0-rc4, it doesn't fail. In fact, it seems like the
issue started happening between 2.0rc6 and 2.0rc7. I'll try and
bisect it down to the commit that introduced the problem, and
hopefully that will provide a better clue.

The reason for this is that busted will reload any requred modules for each spec file. So ffi is being loaded twice. This was done to address issue #62, and has been present since rc5. As a workaround, to prevent this from happening you can put the require in a helper script that is loaded by busted before running any tests.
helper.lua:

local ffi = require 'ffi'

$ busted --helper=helper.lua

or, alternatively, you can do

$ luajit -e 'require "ffi"' /path/to/busted

This should prevent the ffi module from being reloaded by busted.

Maybe LuaJIT doesn't like loading the ffi module multiple times.

The os.exit change, I believe, is new to rc7.

[jszakmeister]

The reason for this is that busted will reload any requred modules for each spec file. So ffi is being loaded twice. This was done to address issue #62, and has been present since rc5. As a workaround, to prevent this from happening you can put the require in a helper script that is loaded by busted before running any tests.

Thanks @o-lim! That does indeed help! Though, I think it might still be worthwhile for LuaJIT to fix though.

[o-lim]

Agreed, it looks like this has revealed some issues in LuaJIT that should be fixed.

@jszakmeister does the workaround fix both the GC and ctype assertion issues or just the ctype assertion issue? Just wondering if the multiple loading of the ffi module causes the GC issue as well, or if they are two separate problems.

[jszakmeister]

@o-lim Yes, both are fixed with the workaround. Thank you so much for helping me out on this!

[DorianGray]

👍

[Tieske]

for reference, a possible fix: lunarmodules/luassert#150