bpo-34070: only check for isatty when buffering < 0

[Dav1dde]

https://bugs.python.org/issue34070

Only check for isatty when buffering actually requires the check (buffering < 0)

https://bugs.python.org/issue34070

[the-knights-who-say-ni]

Hello, and thanks for your contribution!

I'm a bot set up to make sure that the project can legally accept your contribution by verifying you have signed the PSF contributor agreement (CLA).

Unfortunately our records indicate you have not signed the CLA. For legal reasons we need you to sign this before we can look at your contribution. Please follow the steps outlined in the CPython devguide to rectify this issue.

When your account is ready, please add a comment in this pull request
and a Python core developer will remove the CLA not signed label
to make the bot check again.

Thanks again for your contribution, we look forward to reviewing it!

[Dav1dde]

CLA signed.

[serhiy-storchaka]

Would be nice to add tests.

Python implementation in _pyio.py may need a change too.

[Dav1dde]

I'll look into tests.

The Python implementation in Lib/_pyio.py should be fine:

 209   │     try:
 210   │         line_buffering = False
 211   │         if buffering == 1 or buffering < 0 and raw.isatty():
 212   │             buffering = -1
 213   │             line_buffering = True

Since it is left to right. But maybe I am missing something else here?

Regarding CLA: It is signed, I got the copy via E-Mail and my Github account is linked in the Python Bugtracker .. I'll double check it again (maybe it takes a while?).

Also regarding the Misc/News, is that worth a news entry, it's just a rather small bugfix/improvement?

[serhiy-storchaka]

A news entry may save hours for somebody if his program will fail when run on different computer with older Python because of this change.

The Python implementation looks fine to me too.

[vstinner]

I would prefer to have an else block to initialize isatty variable, even if it's not used in practice. ("else { isatty = 0; }"). Or initialize the variable to 0 when it's declared?

It should prevent warning from dummy compilers.

[serhiy-storchaka]

Or add

if (isatty) {
    buffering = 1;
}

and simplify the condition (buffering == 1 || (buffering < 0 && isatty)) below to just (buffering == 1).

[Dav1dde]

Changed. isatty is now initialized and the redundant buffering < 0 below has been removed.

[Dav1dde]

I changed the code you commented on, I turned the now redundant buffering < 0 && isatty into just isatty and assigned 0 as a default to isatty.

I also added a news entry, I hope that is proper English and makes sense for a bystander.

I am still having troubles with writing tests for this though. My initial idea was to replace the FileIO class with my own "spy" like so:

        FileIO = self.io.FileIO

        class SpyFileIO(FileIO):
            _isatty = False

            def isatty(self):
                SpyFileIO._isatty = True
                return False

        self.io.FileIO = SpyFileIO
        try:
            f = self.open(TESTFN, 'w', buffering=-1)
        finally:
            self.io.FileIO = FileIO
            f.close()

        self.assertTrue(SpyFileIO._isatty)

This works fine for the _pyio module, but it does not work for _io - the C module. I assume because the C module doesn't look up the reference from the module for the FileIO type every time (please correct me when I am wrong!). I don't have a better idea on how I could verify the desired behaviour in a test than monkey patch FileIO here. Please point me in the right direction.

[serhiy-storchaka]

LGTM without tests. Thanks @Dav1dde!

Just a small formatting nit.

[serhiy-storchaka]

Use double backquotes:

``buffering=-1``

This is a reStructuredText, not Markdown.

[Dav1dde]

Done!

Sorry about that, haven't been doing enough Python recently :(

[Dav1dde]

Any news on this? Anything I should do?