gh-106505: don't lookup own hostname in getfqdn()

[bluelhf]

• return gethostname() immediately instead of resolving it

• Issue: socket.getfqdn() does gethostbyaddr() lookup for own hostname #106505

[ghost]

All commit authors signed the Contributor License Agreement.

[serhiy-storchaka]

Does it covered by existing tests? If no, please add tests.

Also, please correct the NEWS entry. The sentence should start with a title character and end with a period.

[bluelhf]

@serhiy-storchaka Thanks, I'll fix the NEWS entry. For the tests, though, can you give any advice on how I could implement the test, since the correct result depends on the OS configuration? Is it OK to just call gethostname() and check that the function returns that hostname for empty string and the specified loopback addresses?

[serhiy-storchaka]

I do not know. Does this change affect any of our buildbots?

I would write a PR containing test that compares the result of the old implementation of getfqdn('') with gethostname() and signal if they are different. Also print any interesting details. Then modify some file in Lib/test/libregrtest so that only this test (or only this test file) be run. And run it an buildbots. If there are some interesting results, we can use it to create a test. And it may expose some flaw in this PR, although it does not guaranee this.

[bluelhf]

I do not know. Does this change affect any of our buildbots?

I would write a PR containing test that compares the result of the old implementation of getfqdn('') with gethostname() and signal if they are different. Also print any interesting details. Then modify some file in Lib/test/libregrtest so that only this test (or only this test file) be run. And run it an buildbots. If there are some interesting results, we can use it to create a test. And it may expose some flaw in this PR, although it does not guaranee this.

In many cases, and I assume for all Python buildbots, such tests would pass, as gethostbyaddr would throw an error when passing an invalid address (the computer's hostname), which would cause the hostname to be returned as-is. Sometimes, though, the hostname resolution may take an extended period of time (which is how I originally found this issue, with python -m http.server taking upwards of 10 seconds to start), and since it's incorrect to resolve a hostname as if it were an address anyway, it would be better to just return the hostname without any resolution (as documented!)

For tests, I could write some that mock gethostbyaddr to ensure that it isn't called when an empty name, "::" or "0.0.0.0" are passed to getfqdn()?

[github-actions]

This PR is stale because it has been open for 30 days with no activity.