Re: [SOLVED] Re: and/or missbehaviour in embedded lua 5.1

[Mike Pall <mikelu-0603@...>]

Hi,

Roberto Ierusalimschy wrote:
> Probably we should release a 5.1.1 with it fixed. 5.1.1 can also include
> several small corrections to the manual sent to us after the release of
> 5.1.

Maybe wait a bit until some other things surface?

> We would like to address also the DirectX issue. I sent a message asking
> for suggestions, but got very rew replies...
> 
> - Ok, we should put something in the manual about that. But WHERE? Where
> will someone look in the manual if he gets weird behavior? Maybe a
> "Known (no) Bug" section?

A new FAQ section? Should contain the 'My app crashes when I open
the io library' problem, too (not lua_call'ed.)

> - Should we add some explicit test in the code, too? Maybe a C assert
> in luaH_new?

If the C assert needs to be explicitly turned on (like the other
assertions), then it's worthless. If not then it may be a
performance hit. It also might trigger too late. It triggers when
the first table is created after the init of DirectX, but not
necessarily before the first load/store fails.

Another suggestion: resurrect the lua_number2int MSVC asm macro.
I thought the last iteration of this worked well enough, before
it was replaced by the portable (but precision sensitive) macro.
The gcc asm macros always worked fine.

Bye,
     Mike