version 4.6.1 issue with opener

[yaronbenezra]

Hi,

In the latest version due to "Add support for opener in open() #688" commit we fail to run a tests on python version 3.9 cause:

seems like _OpenModes is now object instead of int

[mrbean-bremen]

Thanks for the report - interesting. _OpenModes has not changed, the difference is that the opener call is now executed here. Looks like the flags have to be converted before passing them into the call - this was not a problem with the tested code, but obvioulsy I didn't test this usage.

Do you have a simple reproducable example, that shows this behavior? This would help to find the problem faster. I have to make another patch release anyway...

[mrbean-bremen]

Never mind, this was a clear mistake. Never a good idea to rush a patch out without adequate tests...

[mrbean-bremen]

Shall be fixed in version 4.6.2.

[yaronbenezra]

i will test and update tomorrow

[yaronbenezra]

Hi
I have tested it and now we fail on :
filesystem.add_real_directory

hope you can help, thanks in advanced

[mrbean-bremen]

Ok, that's unfortunate. I will have a look tonight - it would be nice if you could add a test that can reproduce this.

[yaronbenezra]

I run my test and a good version that all work for me is: 4.5.6
I will try to write a test today so I can upload it for you(I can't expose company code so I need to write a new test)
I saw this :

"changes root path behavior under Windows?" Do I need to change something due to this?

Best Regards,
Yaron

[mrbean-bremen]

"changes root path behavior under Windows?" Do I need to change something due to this?

No, you shouldn't. That just means that internally the handling has changed to better reflect the real behavior under Windows. As this was not a small change, it could have caused some regressions, so it may be related to your problem.
Probably better to open a separate issue for this one...

[mrbean-bremen]

@yaronbenezra - I created the spin-off #691, please add further information to that issue.