Re: [PATCH v28 0/8] MR3870: mspatcha: Add support for 32-bit file patching and normalization

[eric pouech (＠epo)]

eric pouech (@epo) commented about dlls/mspatcha/pa19.c:
> +/* derived from imagehlp for calculating a new coff image checksum. */
> +static WORD calc_chksum(
> - DWORD initial_value, PUCHAR buffer, DWORD size_in_bytes)
>
> +{
> - DWORD sum;
> - DWORD words_remaining;
> - DWORD partial_words;
> - DWORD partial_sum;
> -
> - sum = initial_value;
> - words_remaining = size_in_bytes / sizeof(WORD);
> -
> - for (; words_remaining; sum += HIWORD(partial_sum) + LOWORD(partial_sum))
> - {
> -
>    partial_words = words_remaining;
> -
>    if (words_remaining > 0x10000) {

it could be interesting to explain where the 0x10000 value comes from

and it doesn't look right: each inner loop can add up to sum at most 0x10000 * 0xffff = 0xffff0000, but the value of sum from a previous loop can be >= 0x1000 (because of the lo + hi computation) => you'll overflow the DWORD of sum

reusing tested existing code looks saner to me that reimplementing it