Re: [Bug-wget] Problem with ÅÄÖ and wget

[Ángel González]

On 15/09/13 00:59, Bykov Aleksey wrote:
> Greetings
>
> Great thanks for pushing in correct direction.
>
> With attached patch Wget in Windows can work with UTF-8 names. But - 
> also only with "--restrict-file-names=nocontrol"...
I think there are two issues:
- Make wget recognise utf-8 urls and accept them without nocontrol when 
the filesystem encoding is utf-8.
- Correctly store the filenames in Windows.

I would have started with the first one, and then treat Windows as utf-8 
enabled fs, which is what this patch does. Also, isn't there any library 
doing already this?

> diff --git a/src/utils.c b/src/utils.c
> index 2ec9601..6307c88 100644
> --- a/src/utils.c
> +++ b/src/utils.c
> @@ -2544,3 +2544,42 @@ test_dir_matches_p()
>
>   #endif /* TESTING */
>
> +#ifdef WINDOWS
> +/* For UTF-8 in Windows support. Replacement standart fopen() utime() stat() 
> lstat() mkdir() with wide character
> +analogs route. w_fopen() declared in utils.h, w_utime(), w_stat() and w_mkdir 
> - in utils.c */

This code should be on mswindows.c
What mades w_fopen() different so it is on utils.h instead of the .c?

Commenting on just one function, as they all follow the same templte:

> +int
> +w_stat (const char *filename, struct_stat *buffer )
> +{
> +  wchar_t *w_filename;
> +  int buffer_size = 1024; /* I cant push it to work with strlen() */
What happens if the filename has more than 1024 characters?
> +  w_filename = malloc (buffer_size);
> +  MultiByteToWideChar(65001, 0, filename, -1, w_filename, buffer_size);
Using CP_UTF8 instead of 65001 would be preferable IMHO.

Big bug. The sixth argument is the space available for w_filename *in 
characters*, not bytes.
I would multiply buffer_size by sizeof(wchar_t) in the malloc (although 
you could instead divide here, too).

> +  int res = _wstati64 (w_filename, buffer);
It would be better to declare res at the beginning of the function.

> +  free (w_filename);
> +  return res;
> +}
Why bother allocating memory, if you are using a fixed size? Another 
opiton would be to use alloca()


I guess rename() would also need a wrapper.