Shmuel Zeigerman wrote:
> Jerome Vuarand wrote:
>> Without a dump of the actual data, the problem can come from either
>> struct of FB. Try:
>>
>> a = struct.pack("BHIL",1,2,3,4)
>> print(table.concat({string.byte(a,1,#a)}," "))
>>
>> On my 32bits WinXP it gives:
>>
>> 1 2 0 3 0 0 0 4 0 0 0
>>
>> which is the expected result.
>
> This test on my system gave incorrect result: 1 2 0 3 4 0 0 0 which
> explains why the DLL I sent to Merick didn't work for him (though it
> passed teststruct.lua OK).
>
> The source for the DLL was taken from:
> http://www.inf.puc-rio.br/~roberto/struct/
>
> The problem appears on the line 83 of struct.c, here's the patch:
> - case 'i': return getnum(fmt, sizeof(int));
> + case 'i': case 'I': return getnum(fmt, sizeof(int));
Ah, yes, that's right. I just realized I have such a patch in my
version. It's applied automatically by my build system so I just forgot
about it. I attached my patch file.
The issue was already discussed here in January [1], but the online
version of struct still has the problem.
[1] http://lua-users.org/lists/lua-l/2008-01/msg00439.html
Attachment:
struct-In.patch
Description: struct-In.patch