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

[Tacvek]

----- Original Message ----- 
From: "Ronald Lamprecht" <address@hidden>
To: "Tacvek" <address@hidden>
Cc: "Andreas Lochmann" <address@hidden>; "Daniel 
Heck" <address@hidden>; "enigma-devel" <address@hidden>
Sent: Monday, June 26, 2006 2:25 PM
Subject: Re: [Enigma-devel] Lua 5.1 "luaL_error" problems on Windows


> Hi,
>
> Tacvek wrote:
> > > Some functions supplied by lua.cc crash on calling "luaL_error":
> > >
> > > - enigma.GetAttrib with non existing attribute or pseudo-attribute 
> > > "kind".
> > > - enigma.GetKind(nil)
> > >
> > > These crashes occur only on Windows but not on Linux. They are
> > > independent of Lua 5.1 and 5.1.1.
> > >
> > > The error handling does work in general on Windows, as incorrect calls 
> > > like
> > >
> > > enigma.SetItem(7,9, nil)
> > >
> > > are reported correctly.
> > >
> > > Please find attached a sample level with the different errors commented
> > > out. Activate one after the other for testing.
> > >
> > > Any ideas?
> >
> >
> > Hmm...
> > I can confirm the crash, but my current setup does not support debugging, 
> > because I cross-compile. I could do a direct compilation, and see what 
> > gdb thinks, but It would be a bit of a pain, so I will hold off on doing 
> > so unless nobody else can come up with an explanation.
>
> Looks rather like a Mingw problem:
>
> Trunk r118 (last Lua 4 revision) behaves identical to the current Lua 5.1 
> trunk revision when compiled with the same Mingw gcc 3.4.2. 20040916-1 
> version.
>
> Mingw did change the longjump methods after 3.4.2 and a lot of problems 
> were reported due to principal longjump-Windows incompatibilities. Has 
> anyone experience with candidate 3.4.5?
>
> - Ronald
Yes, that was the version that did not append '.exe' to the names of 
executables.
I'm still not sure if that is intentional.
Since my devel box is a virtual machine, it is easy to test this.

Tested. The 'enigma' executable generated also crashes. So 3.4.5 does not 
seem to be a cure
for this problem.

I find it interesting that this is a clean crash, with the program exiting 
gracefully, and therefore no
Doctor Watson reports are generated.