On 1/25/2012 13:36, Alexandre Julliard wrote:
Nikolay Sivov<[email protected]> writes:
Fix for http://bugs.winehq.org/show_bug.cgi?id=28087
You shouldn't need anything of the kind. If StrStrIW doesn't work right
it should be fixed (plus of course there's no reason to even use
strstr() here at all).
When I debugged that I remember there's was a problem with
CompareStringW() actually.
StrStrIW() eventually uses it like:
---
iRet = CompareStringW(GetThreadLocale(), NORM_IGNORECASE, lpszStr,
iLen, lpszComp, iLen);
---
so both strings are passed with the same length. CompareStringW() in
turn uses winelib's wine_compare_string()
that tries to actualize passed string length look backwards from string
tail searching for first not null - helper real_length().
So it easily gets past string data if specified length is large enough.
Some tests show that CompareStringW()
survives in such case - it uses 'or' it seems, so it doesn't matter what
happens first: a null terminator or consumed chars count.
So yeah, this probably should be fixed, but requires winelib changes and
a test that will crash without a fix, not in very reliable way depending
on string length passed to it.
About avoiding strstr() at all, does this one look ok -
http://bugs.winehq.org/attachment.cgi?id=38556 ?
