gh-118201: Simplify conv_confname

[mhsmith]

On Android and iOS we've been seeing intermittent buildbot failures where os.confstr, os.pathconf or os.sysconf raise a ValueError. This is not a failure in the underlying system call, but a failure to look up a mapping from strings to integers before making the system call.

I haven't been able to explain this (see #118201 for details), but I have found a way to simplify the relevant code, which will hopefully fix the issue, or at least give us some more information about it.

All three functions depend onconv_confname, which currently maps from strings to integers using a binary search in a C array. This PR changes it to use os.{confstr,pathconf,sysconf}_names, which are dicts containing the same information.

I also took the opportunity to improve the test coverage of this area.

• Fixes Intermittent "unrecognized configuration name" failure on iOS and Android #118201
• Reverts gh-118201 - Disable the flaky POSIX test_confstr test on iOS #118452
• Reverts gh-118201: Accomodate flaky behavior of os.sysconf on iOS #118453

[mhsmith]

!buildbot android

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @mhsmith for commit 2a3e50a 🤖

The command will test the builders whose names match following regular expression: android

The builders matched are:

• aarch64 Android PR
• AMD64 Android PR

[mhsmith]

!buildbot ios

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @mhsmith for commit 2a3e50a 🤖

The command will test the builders whose names match following regular expression: ios

The builders matched are:

• iOS ARM64 Simulator PR

[mhsmith]

Well I'm not sure why it ran the buildbots on all platforms when I only asked for two, but I've checked the failures, and all of them are already present on the main branch.

[mhsmith]

@sobolevn: You've done some work on the posix module recently; are you able to review this PR?

[sobolevn]

Nope, sorry, I am not faimiliar with that code :(

[mhsmith]

I'm not sure if any active developer is familiar with this code, but I hope the change should be fairly self-explanatory. @JelleZijlstra: are you able to take a look, or can you recommend anyone else?

[JelleZijlstra]

I can take a look but I'm also not familiar so I'll need a bit of time to familiarize myself with it. Have you tried looking at the git blame for the affected functions to see if any were written by people who are still active?

[mhsmith]

It looks like this code has remained almost unchanged since it was written in 1999 by @freddrake.

[JelleZijlstra]

Wow, that seems like a good reason to be extra careful with any changes.

[mhsmith]

!buildbot android

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @mhsmith for commit a214081 🤖

The command will test the builders whose names match following regular expression: android

The builders matched are:

• AMD64 Android PR
• aarch64 Android PR

[mhsmith]

!buildbot android|ios

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @mhsmith for commit 8a026f1 🤖

The command will test the builders whose names match following regular expression: android|ios

The builders matched are:

• iOS ARM64 Simulator PR
• aarch64 Android PR
• AMD64 Android PR

[JelleZijlstra]

@mhsmith thanks, this looks good! I assume the buildbot failure are the result of the unrelated GC change that came in the other day?

[mhsmith]

Yes, and it looks like that's been fixed on the main branch now, so I'll try another run.

Let's not backport this PR to 3.13, as it's a relatively disruptive change, and a backport isn't really necessary:

• On iOS, the relevant tests are already skipped.
• On Android, the tests have never failed on the 3.13 branch.

[mhsmith]

!buildbot android|ios

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @mhsmith for commit 8a026f1 🤖

The command will test the builders whose names match following regular expression: android|ios

The builders matched are:

• aarch64 Android PR
• iOS ARM64 Simulator PR
• AMD64 Android PR

[mhsmith]

OK, looks like this is ready to merge.