tree 2dcb0d4d4f673512b74d26c27f85a1d86ef1ec11
parent 8fe8b1d829f057c556143a658d3f26bc66c69ee8
author Eli Schwartz <eschwartz93@gmail.com> 1715632063 -0400
committer Eli Schwartz <eschwartz93@gmail.com> 1719173752 -0400
gpgsig -----BEGIN PGP SIGNATURE-----
 
 iQKqBAABCgCUFiEEYEETBMCdNmKDQO7/zrFn77VyK9YFAmZ4gnhfFIAAAAAALgAo
 aXNzdWVyLWZwckBub3RhdGlvbnMub3BlbnBncC5maWZ0aGhvcnNlbWFuLm5ldDYw
 NDExMzA0QzA5RDM2NjI4MzQwRUVGRkNFQjE2N0VGQjU3MjJCRDYWHGVzY2h3YXJ0
 ejkzQGdtYWlsLmNvbQAKCRDOsWfvtXIr1sarD/9pbUzDul/nU2FdXmFyijd4I3Ur
 YJhl1EmP4SqWfs3vOgnYMJq/SzjVvl/LqaRPkAEzyAJJDWBJ424M1R9d6aKWxX+R
 IwBcPUC+6fmOC+o2tnDCCXSnuP+YJu6XeY8I6Ma9RBBFjRrM5hqIMZdWb5g8tV4o
 p9ofeO/gRAIurFkugFOhF9uKX43Ef6Y57vd54UC7u3sRYJOXGcUKlxdu+Le5NBiU
 XnJAXcSddlQy1GqftooVFJ4pby2ot77I5rznMHUrSyOBJG9MmQtS2i1x79JX5S8W
 uBdTvXeUVg37SGEQOzmiDV8YBF0CJwgU/N7KHQf4oQdyOJj8heaatlWmPVVsvzin
 +kmWezLGWuj7ZGBO/hvAlKbMMdN0My6pSAtiNSig9tz6xF49iVFFru+v6eaaojA4
 4wdH4KslpZe2b1P39GfOVMD+WRS8CAqh5aoWOyjtn+wnLaHHSaDv5dH6EhjA87Vh
 y+nvkuaRDuxg4F1JR9WbMFSuR9cYxRMr2DkbO97QixTUujQ32CV7cb8mFUdxOVFL
 cr0Y4yLSO//mVYbjz+m6AWKQpesS8CRRmBphjqnB2acy5Sw3MPY0kV7P/vp9vAAQ
 QIG8BO9i1pUOgqPy/5sxNciua7rLORvAnpN9aECXRVDYOGkEb5BZdRnT7VzckBGZ
 HQyqbFJGqDkW573fXw==
 =qDyi
 -----END PGP SIGNATURE-----

msetup: fix regression under py3.13 causing profile.runctx to not write locals()

"PEP 667: Consistent views of namespaces" caused locals() to be
inconsistent between uses since it is now created afresh every time you
invoke it and writes to it are dropped. `sys._getframe().f_locals` is
equivalent but preserves writes (it doesn't create a new dict) and
unfortunately doesn't help at all as it's documented to be a private
implementation detail of CPython that "should be used for internal and
specialized purposes only".

Work around this by saving locals to a variable reference and both
passing it into runctx and reusing it in lookups of the result. This
works okay for both new and older versions of python.

Per the documentation for locals():

> The contents of this dictionary should not be modified; changes may
> not affect the values of local and free variables used by the
> interpreter.

So... lesson learned? :) This was introduced in commit
c34ee374a77fb2dffff90364506ac0cbbb1f00de; before that, we still used
locals() but only to pass local variables *in*.

Bug: https://github.com/python/cpython/pull/115153