Re: [PATCH] extend svn_subst_translate_string() to record whether re-encoding and/or line ending translation were performed (v. 5)

[Branko Čibej]

On 13.12.2010 12:38, Julian Foad wrote:
> On Fri, 2010-12-10, Danny Trebbien wrote:
> Hi Daniel.
>
> I haven't looked at your patch v5 yet, because this Git diff doesn't
> look quite right.  How will this memcmp expression work properly if the
> two strings being compared are different lengths?
>> -  if (translated_eol != NULL && DIFFERENT_EOL_STRINGS(eol_str, eol_str_len,
>> -                                                      newline_buf, 
>> newline_len))
>> +  if (translated_eol != NULL &&
>> +      memcmp(eol_str,
>> +             newline_buf,
>> +             eol_str_len < newline_len ? eol_str_len : newline_len) != 0)

Not very well, because it's prefix matching that doesn't actually take
different string lengths into account and you can get false matches.
e.g., if eol_str == '\r' (ancient mac mode) and newline_bug == '\r\n'
... what you want is more like:

    if (translated_eol != NULL &&
        (eol_str_len != newline_len ||
         memcmp(eol_str, newline_buff, eol_str_len) != 0))


-- Brane