Re: Patch for relative paths

On Friday, April 25, 2003, at 12:27  PM, Stefano Corsi wrote:

>> I haven't tried the patch just yet but I have questions about the 
>> logic
>>
>>
>> env->meta.filename = moto_strdup(env, filename + strlen(getcwd(cwd,
>> 255)) + 1);
>>
>> How would this work in the case where someone includes (or readdefs) a
>> file from the parent directory with an error in it ?
>
> Eh eh. Forgive me. It was a lazy speedy mess.
>
> Uhm... anyway. I think that we must store somewhere the relative path 
> before
> it gets modified by realpath(). I have tried in struct MetaInfo by 
> adding a
> char * filename_rel field to the struct, but there must be some 
> assumption on
> the MetaInfo struct size, because I get lot of memory errors.
>
I think the above strategy is probably the right one (although we might 
want to name the field orig_filename or something like that since we 
want it to represent exactly the path that was #included or #readdeffed 
in the moto file itself ... I think :) ). My guess is the memory leaks 
came from the filename_rel field contents not being tracked. You can 
probably use motopp_strdup to accomplish this. If you were getting 
memory corruption errors you may need to re-build the entire moto 
directory. There are a couple structures in the parser that do have 
nested MetaInfo structures.

-Dave

> Stefano
>