barry_as_FLUFL future flag does not work in new REPL

[sobolevn]

Bug report

New REPL:

>>> from __future__ import barry_as_FLUFL
>>> 1 <> 2
  File "<python-input-1>", line 1
    1 <> 2
      ^^
SyntaxError: invalid syntax

Old REPL:

>>> from __future__ import barry_as_FLUFL
>>> 1 <> 2
True

I understand that this is just an easter egg :)

Linked PRs

• gh-124960: Fixed barry_as_FLUFL future flag does not work in new REPL #124999
• [3.13] gh-124960: Fixed barry_as_FLUFL future flag does not work in new RE… #125475

[warsaw]

Clearly this is a showstopper bug requiring @Yhg1s to delay Python 3.13.0 for another week and another RC! No respect for the FLUFL eh @pablogsal ??? 😭

[Yhg1s]

I can't believe it, just last week @ambv and @pablogsal were complaining about how easter eggs were getting removed, and here they are, organising a coup! Clearly this should be fixed as soon as possible¹!!!!

Footnotes

1. which means 3.13.1 ↩

[nineteendo]

Hmm, it looks like it's working in the commit that introduced the new REPL: f27f8c7

[nineteendo]

Bisected to a3e4fec.

[nineteendo]

It's failing at this line:

cpython/Lib/_pyrepl/simple_interact.py

Line 86 in a3e4fec

tree = ast.parse(source)

[nineteendo]

Note that this is equivalent to:

tree = compile(source, filename, symbol, ast.PyCF_ONLY_AST)

Which doesn't use codeop.Compile().

[nineteendo]

Is this acceptable?

diff --git a/Lib/_pyrepl/simple_interact.py b/Lib/_pyrepl/simple_interact.py
index d65b6d0d627..6e24860d3c8 100644
--- a/Lib/_pyrepl/simple_interact.py
+++ b/Lib/_pyrepl/simple_interact.py
@@ -83,7 +83,7 @@ def showtraceback(self):
 
     def runsource(self, source, filename="<input>", symbol="single"):
         try:
-            tree = ast.parse(source)
+            tree = self.compile.compiler(source, filename, symbol, ast.PyCF_ONLY_AST)
         except (OverflowError, SyntaxError, ValueError):
             self.showsyntaxerror(filename)
             return False
diff --git a/Lib/codeop.py b/Lib/codeop.py
index 6ad60e7f850..8a77075b487 100644
--- a/Lib/codeop.py
+++ b/Lib/codeop.py
@@ -44,6 +44,7 @@
 # Caveat emptor: These flags are undocumented on purpose and depending
 # on their effect outside the standard library is **unsupported**.
 PyCF_DONT_IMPLY_DEDENT = 0x200
+PyCF_ONLY_AST = 0x400
 PyCF_ALLOW_INCOMPLETE_INPUT = 0x4000
 
 def _maybe_compile(compiler, source, filename, symbol):
@@ -109,11 +110,13 @@ class Compile:
     def __init__(self):
         self.flags = PyCF_DONT_IMPLY_DEDENT | PyCF_ALLOW_INCOMPLETE_INPUT
 
-    def __call__(self, source, filename, symbol, **kwargs):
-        flags = self.flags
+    def __call__(self, source, filename, symbol, flags=0, **kwargs):
+        flags |= self.flags
         if kwargs.get('incomplete_input', True) is False:
             flags &= ~PyCF_DONT_IMPLY_DEDENT
             flags &= ~PyCF_ALLOW_INCOMPLETE_INPUT
+        if flags & PyCF_ONLY_AST:
+            return compile(source, filename, symbol, flags, True)
         codeob = compile(source, filename, symbol, flags, True)
         for feature in _features:
             if codeob.co_flags & feature.compiler_flag:

[pablogsal]

Clearly this is a showstopper bug requiring @Yhg1s to delay Python 3.13.0 for another week and another RC! No respect for the FLUFL eh @pablogsal ??? 😭

[nineteendo]

There's no specific code in the PR to handle this easter egg, so I think the bug applies to all syntax changing futures. It just so happens this is currently the only future that changes syntax. If it did not exist, it might have been very tedious to figure this out when adding one in the future.

Edit: but it's worth noting that old futures changed the parser and not compiler.

[warsaw]

barry_as_FLUFL FTW... again! 🤣

[JelleZijlstra]

@nineteendo's code looks reasonable to me; please submit a PR!

Note that support in the old REPL was also somewhat broken (this is #125331):

% python3.12
Python 3.12.4 (main, Jul 24 2024, 15:24:11) [Clang 15.0.0 (clang-1500.3.9.4)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from __future__ import barry_as_FLUFL; print(1 <> 2)
  File "<stdin>", line 1
    from __future__ import barry_as_FLUFL; print(1 <> 2)
                                                   ^^
SyntaxError: invalid syntax
>>> from __future__ import barry_as_FLUFL
>>> print(1 <> 2)
True
>>> 

[nineteendo]

please submit a PR!

Wulian233 already updated his PR with my patch: #124999.

[ambv]

The immediate fix is landed. Keeping this open because of the long-standing bug preventing the future import to work when parsed in one pass with its usage.

[nineteendo]

Jelle has already opened a separate issue for that: #125331, so I would close this one.

[ambv]

Alright, let's do that then. Thanks!