All commit authors signed the Contributor License Agreement.

Most changes to Pythonrequire a NEWS entry.

Please add it using theblurb_it web app or theblurb command-line tool.

Requires backport to 3.12 (should cleanly apply). Requires backport to 3.11 and 3.10 using this manual cherry-pick:Bo98@b0c571f I believe 3.9 and older don't technically support libedit except Apple's 2012 version, but a similar cherry-pick could be made if desired.

FYI, 3.10 and older is in security-fix-mode; I don't think this applies as a security fix. Backports to 3.12 and 3.11 are allowed for bug-fixes; is this a bugfix?

Most changes to Pythonrequire a NEWS entry.

Please add it using theblurb_it web app or theblurb command-line tool.

Backports to 3.12 and 3.11 are allowed for bug-fixes; is this a bugfix?

I consider it so. It's certainly not a new feature.

Compile error fixes, including those adjusting for new OS and library versions, seem to have been backported before unless I'm mistaken? I consider this the same category as those. Future FreeBSD and NetBSD will ship with this newer editline out of the box too.

Risk should also be low.

Most changes to Pythonrequire a NEWS entry.

Please add it using theblurb_it web app or theblurb command-line tool.

@corona10: you've recently dived into the readline issues. Would you mind taking a look at this PR?

I will take a look by this weekend, I need time to investigate it :)

@corona10: we should have a conditional CI matrix for readline changes, similar to our OpenSSL CI matrix.

Oh, BTW, can you produce a NEWS entry for this,@Bo98?

@corona10: we should have a conditional CI matrix for readline changes, similar to our OpenSSL CI matrix.

Good idea

@erlend-aasland

Requires backport to 3.11 and 3.10 using this manual cherry-pick:Bo98@b0c571f

About the backport patch, I am not sure if we handle this as the bug or not?
IMO, If the user uses CPython <= 3.12, they should use libedit <= 1.53 not the latest version.
How did we deal with similar issue for OpenSSL case?

Since the change was introduced quite recent dayshttp://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libedit/readline/readline.h.diff?r1=1.53&r2=1.54&sortby=date

I think that we have to support old versions(1.53) too.

This PR still supports older versions as it checks for the existence ofrl_compdisp_func_t, which didn't exist prior to 1.54. So < 1.54 will still use the VFunction type cast as before.

This is evidenced by the compile working with Apple's libedit, which is based on 1.33.

About the backport patch, I am not sure if we handle this as the bug or not?
IMO, If the user uses CPython <= 3.12, they should use libedit <= 1.53 not the latest version.
How did we deal with similar issue for OpenSSL case?

Had a dig into this, and found the OpenSSL 3 changes to_hashlib (#84659):

• 3.9 got initial support for the new FIPS API from OpenSSL 3.0 alpha. 3.8 didn't receive a backport for this change as 3.8 didn't contain FIPS API so there was no need to backport. 3.8 did receive a backport to satisfy the compiler and silence deprecation warnings (OpenSSL 3.0.0: define OPENSSL_API_COMPAT 1.1.1 #87965).
• Final release of OpenSSL 3.0 came when 3.11 was themain branch, but the fixes for the final release were backported to 3.10 and 3.9.
• _ssl changes were split into multiple issues, but seemed to be backported accordingly at the time, e.g.OpenSSL 3.0.0: handle empty cadata consistently #88086.

Some of those backports (e.g._hashopenssl.c in4ddd5da) were fairly large too, and could have affected runtime behaviour. Tbh, I would normally have expected the bar to be higher to only cover compile-level adjustments or small changes, though it seems that wasn't the case here.

With that said, the fixes for libedit in this PR are only compile-level adjustments - there is no runtime change as it's just a type cast adjustment. Both branches of the#if defined are identical otherwise.

I think that we have to support old versions(1.53) too. WDYT?

It seems to be there is no problem with old versions;AC_CHECK_TYPE will check for the type def and the correct signature will be used.

About the backport patch, I am not sure if we handle this as the bug or not?

Build failure => build bug. IMO, it is a bug fix.

IMO, If the user uses CPython <= 3.12, they should use libedit <= 1.53 not the latest version.

Why?

If we are to restrict the upper version requirement for a third-party dependency, we must document that, and preferrably also implement compile time and runtime checks for that. I don't think any other dependency has an upper version limit.

How did we deal with similar issue for OpenSSL case?

Which issue were you thinking of?

Thanks@Bo98 for the PR, and@corona10 for merging it 🌮🎉.. I'm working now to backport this PR to: 3.11, 3.12.
🐍🍒⛏🤖

Sorry,@Bo98 and@corona10, I could not cleanly backport this to3.11 due to a conflict.
Please backport usingcherry_picker on command line.

cherry_picker f4cb0d27cc08f490c42a22e646eb73cc7072d54a 3.11

GH-110562 is a backport of this pull request to the3.12 branch.

Sorry,@Bo98 and@corona10, I could not cleanly backport this to3.11 due to a conflict.

@Bo98, do you want to do the backport?

GH-110575 is a backport of this pull request to the3.11 branch.