In 'lundump.c', when loading the upvalues of a function, there can be
a read error if the chunk is truncated. In that case, the creation
of the error message can trigger an emergency collection while the
prototype is still anchored. So, the prototype must be GC consistent
before loading the upvales, which implies that it the 'name' fields
must be filled with NULL before the reading.
'lua_resetthread' should reset the CallInfo list before calling
'luaF_close'. luaF_close can call functions, and those functions
should not run with dead functions still in the CallInfo list.
Avoid undefined behavior in calls like «fprintf("%s", NULL)».
('lua_writestringerror' is implemented as 'fprintf', and 'lua_tostring'
can return NULL if object is not a string.)
The parser were mixing compiler indices of variables with stack indices,
so that when a to-be-closed variable was used inside the scope of
compile-time constants (which may be optimized away), it might be closed
in the wrong place. (See new tests for examples.)
Besides fixing the bugs, this commit also changed comments and variable
names to avoid that kind of confusion and added tests.
- more consistent nomenclature for error handling
- more precise definition for dead objects
- added algorithm used by 'math.random'
- added luaL_pushfail
- some other minor changes
ISO C is silent about the return of 'system'. Windows sets 'errno' in
case of errors. Linux has several different error cases, with different
return values. ISO C allows 'system' to set 'errno' even if there are no
errors. Here we assume that a status==0 is success (which is the case
on several platforms), otherwise it is an error. If there is an error
number, gives the error based on it. (The worst a spurious 'errno'
can do is to generate a bad error message.) Otherwise uses the normal
results.
(Undoing part of commit f53eabeed8.) It is better to keep this encoding
stable, so that all Lua versions can read at least the version of a
binary file.
gcc now warns (with -Wextra) about casts between pointers to different
function types. The type 'void(*)(void)' works as a 'void*' for function
pointers, cleaning the warning.
The code should not compute an instruction address before checking that
it exists. (Virtually no machine complains of computing an invalid
address, as long as the address is not used, but for ISO C that is
undefined behavior.)
'simplesect' encloses the introductory text of sections with
subsections, so that each section either is all text or is all
subsections. (This commit also corrects a small brace error in the
manual and extra spaces/tabs in some other files.)
Collisions in short strings occurr just by their existence, when
internalizing them. (Collisions in long strings is caused/controlled
by the program, when adding them as keys to the same table.)
This function was computing invalid instruction addresses when the
expression was not a multi-return instruction. (Virtually all machines
don't raise errors when computing an invalid address, as long as the
address is not accessed, but this computation is undefined behavior in
ISO C.)
Instead of an explicit value (field 'b'), true and false use different
tag variants. This avoids reading an extra field and results in more
direct code. (Most code that uses booleans needs to distinguish between
true and false anyway.)
ISO C states that standard library functions should not be called
with NULL arguments, unless stated otherwise. 'sprintf' does not
state otherwise, and it doesn't hurt to be on the safe side.
