C code:

union {
    struct {
        void *next;
        void *prev;
    } _gc;
    long double alignment;
} PyGC_Head;

int main()
{
    return sizeof(PyGC_Head);
}

With long double:

vstinner@apu$ gcc x.c -o x && ./x; echo $? # 64-bit void*
16
vstinner@apu$ gcc -m32 x.c -o x && ./x; echo $? # 32-bit void*
12

Without long double:

vstinner@apu$ gcc x.c -o x && ./x; echo $? # 64-bit void*
16
vstinner@apu$ gcc -m32 x.c -o x && ./x; echo $? # 32-bit void*
8

I'm not sure that 12 bytes is great for alignment. 32-bit CPU may be more efficient on SIMD instructions with 8 bytes alignment, no?

@methane also proposed PR #13336 for master, but he rejected it.

PyGC_Head in master branch is two uintptr_t. It is 8 bytes (not 12 bytes) on x86.

I'm not sure that 12 bytes is great for alignment. 32-bit CPU may be more efficient on SIMD instructions with 8 bytes alignment, no?

This pull request changes from 12 bytes to 12 or 16 bytes, no?

Sorry @methane, I don't understand this change :-( What is the effect of PyListObject for example? Is the address of ob_item aligned to 16 bytes on x86 (32-bit)? Is it on x86-64 (64 bit)?

@vstinner You wrote:

    struct {
        void *next;
        void *prev;
    } _gc;

But it is actually:

    struct {
        void *next;
        void *prev;
        ssize_t gc_refs; /* only in Python < 3.8 */
    } _gc;

This is very big difference between Python 3.7 and 3.8. That's why I closed PR for 3.8 but don't close this.

sizeof(PyGC_Head) is 12 byte on x86, regardless alignment is double or long double. (I can now "gcc -m32" on Ubuntu, thank you)
On amd64, sizeof(PyGC_Head) is 24 byte with double alignment, and 32byte with long double alignment.

What is the effect of PyListObject for example? Is the address of ob_item aligned to 16 bytes on x86 (32-bit)? Is it on x86-64 (64 bit)?

This pull request affects only where PyObject* ((PyObject*)(gc + 1)) is aligned.
This pull request doesn't affect alignment in PyListObject. offsetof(PyListObject, ob_item) is 24.

⚠️⚠️⚠️ Buildbot failure ⚠️⚠️⚠️

Hi! The buildbot AMD64 Debian PGO 3.7 has failed when building commit ea2b76b.

What do you need to do:

1. Don't panic.
2. Check the buildbot page in the devguide if you don't know what the buildbots are or how they work.
3. Go to the page of the buildbot that failed (https://buildbot.python.org/all/#builders/128/builds/1189) and take a look at the build logs.
4. Check if the failure is related to this commit (ea2b76b) or if it is a false positive.
5. If the failure is related to this commit, please, reflect that on the issue and make a new Pull Request with a fix.

You can take a look at the buildbot page here:

https://buildbot.python.org/all/#builders/128/builds/1189

From https://github.com/python/cpython
 * branch                  3.7        -> FETCH_HEAD
Reset branch '3.7'

find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
make[2]: [clean] Error 1 (ignored)
Segmentation fault
make[3]: *** [pybuilddir.txt] Error 1
make[2]: *** [build_all_generate_profile] Error 2
make[1]: *** [profile-gen-stamp] Error 2
make: *** [profile-run-stamp] Error 2

find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
make: [clean] Error 1 (ignored)

⚠️⚠️⚠️ Buildbot failure ⚠️⚠️⚠️

Hi! The buildbot AMD64 Debian PGO 3.7 has failed when building commit ea2b76b.

What do you need to do:

1. Don't panic.
2. Check the buildbot page in the devguide if you don't know what the buildbots are or how they work.
3. Go to the page of the buildbot that failed (https://buildbot.python.org/all/#builders/128/builds/1190) and take a look at the build logs.
4. Check if the failure is related to this commit (ea2b76b) or if it is a false positive.
5. If the failure is related to this commit, please, reflect that on the issue and make a new Pull Request with a fix.

You can take a look at the buildbot page here:

https://buildbot.python.org/all/#builders/128/builds/1190

Reset branch '3.7'

find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
make[2]: [clean] Error 1 (ignored)
Segmentation fault
make[3]: *** [pybuilddir.txt] Error 1
make[2]: *** [build_all_generate_profile] Error 2
make[1]: *** [profile-gen-stamp] Error 2
make: *** [profile-run-stamp] Error 2

find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
find: ‘build’: No such file or directory
make: [clean] Error 1 (ignored)

this change needs to be undone. followup on the bug.

Do you think missing "build" directory is happened by this change?

I thought it just a trouble on the build bot environment.

The resulting python interpreter was crashing on the the PGO bot and a pile of new undefined behavior was happening on the undefined behavior sanitizer (ubsan) bot.

PGO bot:

./python -E -S -m sysconfig --generate-posix-vars ;\
if test $? -ne 0 ; then \
	echo "generate-posix-vars failed" ; \
	rm -f ./pybuilddir.txt ; \
	exit 1 ; \
fi
Segmentation fault
generate-posix-vars failed
Makefile:605: recipe for target 'pybuilddir.txt' failed

undefined behavior sanitizer millions of times:

Objects/funcobject.c:663:5: runtime error: member access within misaligned address 0x7fb4b5a03a08 for type 'union _gc_head', which requires 16 byte alignment
0x7fb4b5a03a08: note: pointer points here
 b4 7f 00 00  70 8d c6 b5 b4 7f 00 00  a0 13 a0 b5 b4 7f 00 00  00 00 00 00 00 00 00 00  00 00 00 00
              ^ 
Objects/funcobject.c:663:5: runtime error: member access within misaligned address 0x7fb4b5a03a08 for type 'struct (anonymous struct at ./Include/objimpl.h:253:5)', which requires 16 byte alignment
0x7fb4b5a03a08: note: pointer points here
 b4 7f 00 00  70 8d c6 b5 b4 7f 00 00  a0 13 a0 b5 b4 7f 00 00  00 00 00 00 00 00 00 00  00 00 00 00
              ^ 
Objects/funcobject.c:663:5: runtime error: store to misaligned address 0x7fb4b5a03a08 for type 'union _gc_head *', which requires 16 byte alignment
0x7fb4b5a03a08: note: pointer points here
 b4 7f 00 00  70 8d c6 b5 b4 7f 00 00  a0 13 a0 b5 b4 7f 00 00  00 00 00 00 00 00 00 00  00 00 00 00
              ^ 

#13318 fixed it already.