On 3/23/2010 16:21, Dan Kegel wrote:
Hi Eric, welcome to Wine! http://msdn.microsoft.com/en-us/library/ms221240.aspx says the only time those two functions disagree on valid bstr's is if there are embedded nul chars, in which case SysStringLength probably returns a larger value.
Hi Dan.
It's a quite obvious thing. You can use BSTR in place of any WCHAR based parameter, keeping in mind that nul inside BSTR will be treated as end of string, and BSTR itself is always nul-terminated. In this case embedded nulls aren't allowed for node names I suppose, at least http://www.w3.org/TR/REC-xml/#NT-Name tells that it's not possible.
The problem here is that native implementation most likely allows this parameter to be WCHAR string, and this is a case for ::loadXML() method (already fixed). It's not a big deal to fix it, with tests of course.
Are you quite, quite sure that your bstrs are valid?
Most likely they are not, which indicates an application bug or our BSTR management deficiency, I believe there was a discussion in bugzilla about some caching functionality for example we don't provide, not sure it could be related, just as an example.
Eric, please show us a code you're using to create BSTR before passing it to msxml.
(In general, the best way to answer questions like that is to add a test case to your patch, and make sure it passes on Windows as well as Wine before submitting to wine-patches.)
I already asked for that.
- Dan