Winetest: What to do if NT4 behaves other then W2k and up

[Uwe Bonnes]

Wine Bug 21292 shows a problem with
CreateFileA("bla/n", GENERIC_READ, FILE_SHARE_READ | FILE_SHARE_WRITE,
 NULL, OPEN_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL)

All winebot systems beside WNT4WSSP6 return ERROR_INVALID_NAME 123L, while
wine and NT4 happily open the file. How to proceed?

=== WNT4WSSP6 (NT4 Workstation SP6 (English, IE2)) ===
file.c:875: Test failed: CreateFileA failed on nonexist.ent/, hFile
000000B4, err=0, should be 123

=== W2KPROSP4 (Windows 2000 Professional SP4) ===
No test failures found

=== WXPPROSP3 (XP Professional SP3 (English, IE6)) ===
No test failures found

=== W2K3R2SESP2 (Server 2003 R2 Standard Edition SP2) ===
No test failures found

=== WVISTAADM (Windows Vista (English, IE7)) ===
No test failures found

=== W2K8SE (Server 2008 Standard Edition) ===
No test failures found

[Austin English]

On Sat, Jan 16, 2010 at 1:06 PM, Uwe Bonnes
 wrote:
> Wine Bug 21292 shows a problem with
> CreateFileA("bla/n", GENERIC_READ, FILE_SHARE_READ | FILE_SHARE_WRITE,
>  NULL, OPEN_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL)
>
> All winebot systems beside WNT4WSSP6 return ERROR_INVALID_NAME 123L, while
> wine and NT4 happily open the file. How to proceed?

Sounds like you should use the broken() macro...

[Paul Vriens]

On 01/17/2010 01:35 AM, Austin English wrote:
> On Sat, Jan 16, 2010 at 1:06 PM, Uwe Bonnes
>   wrote:
> > Wine Bug 21292 shows a problem with
> > CreateFileA("bla/n", GENERIC_READ, FILE_SHARE_READ | FILE_SHARE_WRITE,
> >   NULL, OPEN_ALWAYS, FILE_ATTRIBUTE_NORMAL, NULL)
> >
> > All winebot systems beside WNT4WSSP6 return ERROR_INVALID_NAME 123L, while
> > wine and NT4 happily open the file. How to proceed?
>
> Sounds like you should use the broken() macro...

broken() won't work as the test does not succeed on Wine.

Just to have the tests pass we need a todo_wine() and a broken() 
construct. I do however think that Uwe is talking about how to handle 
our implementation.

I've tested your patch and Win9x/WinMe are fine. I'd say if only NT4 is 
wrong here we should mark NT4 broken() in the tests and adjust our 
implementation to follow the XP results.