Re: [Enigma-devel] Lua 5.1 "luaL_error" problems on Windows

[Ronald Lamprecht]

Hi,

Tacvek wrote:
> > > > > It would also explain the consistancy. The quickest way to tell is to
> > > > > use --disable-nls.
> >
> > Btw. Could somebody check in the fixes for --disable-nls? In one file 
> > the output of _() is assigned to a char* when a "const char*" is what 
> > is needed. The adding the const keyword in two places should work, I 
> > know const_cast<char*> works but it is a c++ string, which is not 
> > supposed to be mutable. In othe other place, somebody accidentally 
> > used gettext() trather than _(). Both are trivial fixes.

I will check them lateron and submit the fixes with my next commit.

> > > > > I tested that. Besides a few small tweaks that are needed to make it
> > > > > compile
> > > > > using with that, it worked.
> > > > > The resulting exe also crashed.
> > > > >
> > > > > So my next guess is an uncaught C++ exception.
> > > >
> > > >
> > > > A few weeks ago I managed to compile Enigma using the Visual Studio 
> > > > Express.  I have not checked in all the changes that were necessary, 
> > > > but I can try to reproduce this error to see if it is actually a 
> > > > mingw bug. I will be back in Germany on Saturday, so maybe I can 
> > > > take a look at it during the weekend.
> > >
> > > This crosscheck would be helpfull.
> > >
> > > Meanwhile I succeeded in running Enigma on Windows under the control of
> > > gdb (still just with the commandline interface).
> >
> >
> > Well, I found out that the functions that use lua_tostring will 
> > normally crash if they emit an error.
>
>
> Hmm... this seems to be limited to the callback. The other functions 
> seems to act fine in the main body.
>
> Its starting to look like the callback and the main body bugs are quite 
> different.
>
> After all, the unoptimised exe will not crash from getkind(nil) in the 
> body, but will
> crash with that in the callback.

I had a look in gdb on the "enigma.GetKind(nil)" case - it just causes a 
segmentation fault at the same line 166 in lobject.c.

In both cases immediatly before a luaM_realloc_ occured on L. That may 
be the common internal reason.

If someone likes to view the step protocol of the "enigma.GetKind(nil)" 
case please send me a personal mail.

- Ronald