Never seen this before on a build agent:

 /io/buildconfig/manylinux-build/build-wheels.sh: line 60:  6067 Segmentation fault      (core dumped) ${PYTHON} -m pygame.tests -vv --exclude opengl,music,timingMakefile:13: recipe for target 'wheels-manylinux2010-x86' failedmake: *** [wheels-manylinux2010-x86] Error 1Error: Process completed with exit code 2.

Looking deeper intocolor, it does:

#defineCOLOR_TPFLAGS_COMMON \    (Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_CHECKTYPES)#defineCOLOR_TPFLAGS COLOR_TPFLAGS_COMMON

And then the only place it usesCOLOR_TPFLAGS_COMMON orCOLOR_TPFLAGS is the one place in the tp_flags field.

Additionally,Py_TPFLAGS_CHECKTYPES isn't even in the docs anymore. I could only find it in the 2.7 docs, so it's probably just antiquated and doing nothing now. Although testing would be needed.

Yeah I believePy_TPFLAGS_CHECKTYPES can be safely removed now. Because it is defined to0 in a compat header. Also I'll make a commit simplifying alltp_flags everywhere

EDIT: Apparently thosePy_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE flags need to be explicitly passed even though they are the default

The latest commit adds static dot declarations for more structs, mainly for consistency reasons.
The rationale behind this PR is to make these structs easier to maintain in the long run, and make it less error prone, because the current system heavily relies on maintaining the right order of stuff. I believe these new struct declarations will also make it easier for new contributors who are more familiar with the same. It also makes the code more concise and easier to understand IMHO

But we still need to check for whether this can break on any old compiler on any exotic configuration, which we support (I don't think this is the case because this PR seems to pass on CI, the fail ATM is unrelated)