[cfe-dev] [libcxx][windows]

[Howard Hinnant]

I received the following suggestions concerning src/support/win32/support.cpp.  Could someone with a stake in libcxx on Windows take a look at this and make a recommendation? Thanks.

------------
The nature of the fixes means it certainly wont make things worse, though I imagine the whole file will get deleted eventually.
 
The first bug is on line 33 of support.cpp. See the comment.
int vasprintf( char **sptr, const char *__restrict fmt, va_list ap )
{
    *sptr = NULL;
    int count = vsnprintf( *sptr, 0, fmt, ap );
    if( (count >= 0) && ((*sptr = (char*)malloc(count+1)) != NULL) )
    {
        vsprintf( *sptr, fmt, ap );
   /// -- The line below is wrong and needs to be removed as I've done here.
   /// spr[count] refers to a pointer not a character and vsprintf will null terminate anyway.
   /// Left in, it will scribble over random memory quite a ways from the source.
        //sptr[count] = '\0'; // <--- WRONG
    }
 
return count;
}
 
The second bug is on line 60 of support.cpp: I've no idea about the code in general, but the new should be +1. see the comment.
size_t wcsnrtombs( char *__restrict dst, const wchar_t **__restrict src,
                   size_t nwc, size_t len, mbstate_t *__restrict ps )
{
 // <---- The line below should be nwc+1.
    wchar_t* local_src = new wchar_t[nwc];
    wchar_t* nwcsrc = local_src;
    wcsncpy(nwcsrc, *src, nwc);
 // <----- The line below is fine buy definite overwrite here without adding +1 above.
    nwcsrc[nwc] = '\0';
    const size_t result = wcsrtombs( dst, const_cast<const wchar_t **>(&nwcsrc), len, ps );
    // propogate error
    if( nwcsrc == NULL )
        *src = NULL;
    delete[] nwcsrc;
    return result;
}
------------

Howard