gh-115999: Disable the specializing adaptive interpreter in free-threaded builds

[swtaarrs]

For now, disable all specialization when the GIL might be disabled. We can reenable functionality as it is properly supported.

Disabling specialization was just a few lines in pycore_code.h, then there are two related changes:

• Define an alternate implementation of DECREMENT_ADAPTIVE_COUNTER() for free-threaded builds that aborts, as a sanity check.

• Add some code to the interpreter generator to avoid unused variable warnings for variables only used by code within #ifdef ENABLE_SPECIALIZATION blocks (like this_instr and counter here). There are a variety of ways to suppress these warnings. I went with (void)foo; because it seemed like the most straightforward option, but I'm open to other suggestions. (Py_UNUSED(foo) won't work since it's intended for variables that are completely unused, and it changes the variable's name to _unused_foo.)

• Issue: Make the specializing interpreter thread-safe in --disable-gil builds #115999

[colesbury]

!buildbot nogil

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @colesbury for commit db4f433 🤖

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

The builders matched are:

• x86-64 MacOS Intel ASAN NoGIL PR
• x86-64 MacOS Intel NoGIL PR
• ARM64 MacOS M1 Refleaks NoGIL PR
• ARM64 MacOS M1 NoGIL PR
• AMD64 Ubuntu NoGIL Refleaks PR
• AMD64 Ubuntu NoGIL PR
• AMD64 Windows Server 2022 NoGIL PR

[colesbury]

This looks good to me, but I'd like someone working on the specializing interpreter to review it as well.

@brandtbucher, I saw your comment #115999. Would you be able to review this?

[gvanrossum]

I'll do the review. Looks good to me too. I was wondering if DECREMENT_ADAPTIVE_COUNTER() shouldn't just be an error. Otherwise only 1-2 nits.

[gvanrossum]

Let me know if you need help figuring out the remaining test failures.

[swtaarrs]

The test_interpreters refleak failure happens intermittently on main with a normal build for me. I'll see what I can figure out.

[colesbury]

@swtaarrs - it looks like the free-threaded refleaks builds are broken on main:

https://buildbot.python.org/all/#/builders/1226
https://buildbot.python.org/all/#/builders/1368

[swtaarrs]

Yeah, I've been trying (unsuccessfully) to bisect the failure for a bit. I'll update this PR with Guido's comment soon either way.

[swtaarrs]

This should be ready for review again. The test failures are from gh-116099.

[gvanrossum]

Yeah! After the other PR lands you can merge the new main and then maybe @colesbury can merge this? (Or set up auto-merge.)