bpo-33608: Normalize atomic macros so that they all expect an atomic struct#12877
bpo-33608: Normalize atomic macros so that they all expect an atomic struct#12877zooba merged 1 commit into
Conversation
|
Looks good. It makes sense to me that the macro name implies the argument type. |
Sorry, something went wrong.
|
I triggered a custom buildbot run to validate this across the fleet. |
Sorry, something went wrong.
|
The custom buildbots seem happy enough, so I'll give Eric some time to review if he wants and then I'll merge. |
Sorry, something went wrong.
ericsnowcurrently
left a comment
There was a problem hiding this comment.
The approach is good. It definitely reads better. :) I only had a chance to skim through the PR. If you'd like me to dig in and verify each line (which could be valuable given the sensitivity of this code) then I'll see about that tomorrow afternoon. Otherwise LGTM.
Sorry, something went wrong.
|
Mostly wanted an "OK" on the approach. The compiler is doing the most important checks |
Sorry, something went wrong.
@ericsnowcurrently @paulmon This should be an alternative fix to #12665.
Basically, any macro/function starting with
_Py_atomicnow always requires a reference to a_Py_atomicstruct, and never the extracted value.https://bugs.python.org/issue33608