On Thu, 2006-09-28 at 20:44 -0700, Andrew Bartlett wrote: > On Thu, 2006-09-28 at 18:53 -0400, simo wrote: > > On Thu, 2006-09-28 at 17:06 +0000, [EMAIL PROTECTED] wrote: > > > Author: abartlet > > > Date: 2006-09-28 17:06:38 +0000 (Thu, 28 Sep 2006) > > > New Revision: 18978 > > > > > > WebSVN: > > > http://websvn.samba.org/cgi-bin/viewcvs.cgi?view=rev&root=samba&rev=18978 > > > > > > Log: > > > Fix bug found by: > > > http://www.ee.oulu.fi/research/ouspg/protos/testing/c06/ldapv3/ > > > > > > The issue here is that if the UTF8 conversion fails, because this > > > isn't actually UTF8 data, then we need to do a binary compare instead. > > > > I think we should just fail. Why should we compare wrong data anyway? > > Can you give me a valid case where we want to allow invalid utf8 > > strings? > > Imagine a qsort() function, based on this comparison: What would happen > if two strings always returned '-1' against each other, even if > reversed?
They are invalid utf8 strings, we must not have them on in our tree in the first place. If the user sends in an invalid utf8 string for comparison I don't care either, it's the user fault. > I'm not sure there is a valid way to fail this, and given this is the > default comparison function, a binary compare seems reasonable to me... No, the more I think about it the more I think we should fail the comparison, and we also need a way to report failure probably. Simo. -- Simo Sorce Samba Team GPL Compliance Officer email: [EMAIL PROTECTED] http://samba.org
