it is possible for someone to mess with the global table in the prelude or lua_init phases of thread creation. For instance, one could disable the creation of global variables via the _G __newindex metatable method. Then the thread spawning code would not work correctly. For this reason we should not rely on normal insertion into _G and instead use rawset
Coverage decreased (-0.03%) to 64.52% when pulling 322ab8627a8aa61a58938f9df945d1fdacf83136 on johnfoconnor:master into e1ca8e166f867a39960ec8ce6d5db00a9728c1db on zeromq:master.
Coverage decreased (-0.03%) to 64.52% when pulling 322ab8627a8aa61a58938f9df945d1fdacf83136 on johnfoconnor:master into e1ca8e166f867a39960ec8ce6d5db00a9728c1db on zeromq:master.
Coverage remained the same at 64.55% when pulling 322ab8627a8aa61a58938f9df945d1fdacf83136 on johnfoconnor:master into e1ca8e166f867a39960ec8ce6d5db00a9728c1db on zeromq:master.
it is possible for someone to mess with the global table in the prelude or lua_init phases of thread creation. For instance, one could disable the creation of global variables via the _G __newindex metatable method. Then the thread spawning code would not work correctly. For this reason we should not rely on normal insertion into _G and instead use rawset