Idea of fixing SHFileOperation(FO_MOVE)

[Zhenbo Li]

Hello,

As I'm working on bug 25207, I found that previous test code
(test_move() in shlfileop.c) had some disadvantages

1. Duplicate tests
2. Legacy test code (for Win98 or NT4)
3. Lack of test for Vista or later Windows Versions

As more applications are depending on the new behaviours introduced in
Vista, IMHO, we should focus on Vista's new behaviour. So I rewrote
test_move() to remove code for win98, mark xp behaviour as broken, and
expect Vista's behaviour. My next step is to re-implement FO_MOVE
operation of SHFileOperation to fix bug 25207.

Any suggestion for this?
Thank you.

[Bruno Jesus]

On Thursday, September 3, 2015, Zhenbo Li  wrote:
> Hello,
>
> As I'm working on bug 25207, I found that previous test code
> (test_move() in shlfileop.c) had some disadvantages
> - Duplicate tests
> - Legacy test code (for Win98 or NT4)
> - Lack of test for Vista or later Windows Versions
>
> As more applications are depending on the new behaviours introduced in
> Vista, IMHO, we should focus on Vista's new behaviour. So I rewrote
> test_move() to remove code for win98, mark xp behaviour as broken, and
> expect Vista's behaviour. My next step is to re-implement FO_MOVE
> operation of SHFileOperation to fix bug 25207.
>
> Any suggestion for this?

Hi, I dislike the idea of removing tests because they are obsolete but that
is just my opinion. At least I like to keep winsock running in NT4.

What I have done in patches that reached similar issues is to detect the OS
using a specific test that you are sure differs between OS and then use
if/else to separate the tests. This reduces the amount of broken required,
broken can be a vilain sometimes because it does not specify which OS is
failing.

Best wishes,
Bruno

[Zhenbo Li]

Hi Bruno,

Thanks for your comments.

2015-09-03 16:04 GMT+08:00 Bruno Jesus :
> Hi, I dislike the idea of removing tests because they are obsolete but that
> is just my opinion. At least I like to keep winsock running in NT4.

Do you know how I can get the test result in NT4? I have no idea about that.
> What I have done in patches that reached similar issues is to detect the OS
> using a specific test that you are sure differs between OS and then use
> if/else to separate the tests. This reduces the amount of broken required,
> broken can be a vilain sometimes because it does not specify which OS is
> failing.

How about this soultion?
ok(retval == value1 || broken(retval == value2) || broken(retval == value3) ...)
switch(retval)
{
case value1:...
case value2:...
case value3:...
}

In this way, we can find the difference between OSs.

[Bruno Jesus]

On Thu, Sep 3, 2015 at 10:17 PM, Zhenbo Li  wrote:
> Hi Bruno,
>
> Thanks for your comments.
>
> 2015-09-03 16:04 GMT+08:00 Bruno Jesus :
> > Hi, I dislike the idea of removing tests because they are obsolete but that
> > is just my opinion. At least I like to keep winsock running in NT4.
>
> Do you know how I can get the test result in NT4? I have no idea about that.

Not really, I run tests manually. I don't think you should bother about that.
> > What I have done in patches that reached similar issues is to detect the OS
> > using a specific test that you are sure differs between OS and then use
> > if/else to separate the tests. This reduces the amount of broken required,
> > broken can be a vilain sometimes because it does not specify which OS is
> > failing.
>
> How about this soultion?
> ok(retval == value1 || broken(retval == value2) || broken(retval == value3) ...)
> switch(retval)
> {
> case value1:...
> case value2:...
> case value3:...
> }
>
> In this way, we can find the difference between OSs.

What I was thinking was more simple like:

retval = test_with_known_os_difference();
if(retval = VISTA_RESULT)
{
   //all >= vista tests
}
else
{
  //all xp + old broken tests
}

Others may have different opinion, this is just the way I do =)