obj: add footer ignoring for valgrind

[osalyk]

https://github.com/daos-stack/pmdk/actions/runs/20145268034

This change is 

[janekmi]

@janekmi reviewed all commit messages.
Reviewable status: 0 of 1 files reviewed, 1 unresolved discussion (waiting on @grom72 and @osalyk)

---

src/libpmemobj/memblock.c line 1185 at r1 (raw file):

	*(hdr + size_idx - 1) = f;
	/* no need to persist, footers are recreated in heap_populate_buckets */
	VALGRIND_SET_CLEAN(hdr + size_idx - 1, sizeof(f));

Do we still need this one?

[osalyk]

Reviewable status: 0 of 1 files reviewed, 1 unresolved discussion (waiting on @grom72 and @janekmi)

---

src/libpmemobj/memblock.c line 1185 at r1 (raw file):

Previously, janekmi (Jan Michalski) wrote…

Do we still need this one?

Yes.

[grom72]

@grom72 reviewed 1 of 1 files at r1, all commit messages.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @janekmi)

[grom72]

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @janekmi)

[janekmi]

@janekmi reviewed 1 of 1 files at r1.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @osalyk)

---

src/libpmemobj/memblock.c line 1185 at r1 (raw file):

	*(hdr + size_idx - 1) = f;
	/* no need to persist, footers are recreated in heap_populate_buckets */
	VALGRIND_SET_CLEAN(hdr + size_idx - 1, sizeof(f));

1. It would be better keep two "ignore" calls together and have a common description for both of them. But I have a suspicion the TX-related call has to be done before the memory is modified, hasn't it? 😉
2. The comment you proposed sounds a lot more sophisticated but I could not find the source of the terms you used and I do not want to introduce new terms like "persistent storage" or "transactional storage". It is very important to use one vocabulary across the codebase. The comment I propose below is a simple extension of the comment which was already here. But I am happy to discuss it further.

Suggestion:

	*(hdr + size_idx - 1) = f;
	/*
	 * no need to add to a transaction nor persist, footers are recreated in
	 * heap_populate_buckets
	 */
	VALGRIND_ADD_TO_GLOBAL_TX_IGNORE(hdr + size_idx - 1, sizeof(f));
	VALGRIND_SET_CLEAN(hdr + size_idx - 1, sizeof(f));

[osalyk]

Reviewable status: 0 of 1 files reviewed, 1 unresolved discussion (waiting on @grom72 and @janekmi)

---

src/libpmemobj/memblock.c line 1185 at r1 (raw file):

Previously, janekmi (Jan Michalski) wrote…

1. It would be better keep two "ignore" calls together and have a common description for both of them. But I have a suspicion the TX-related call has to be done before the memory is modified, hasn't it? 😉
2. The comment you proposed sounds a lot more sophisticated but I could not find the source of the terms you used and I do not want to introduce new terms like "persistent storage" or "transactional storage". It is very important to use one vocabulary across the codebase. The comment I propose below is a simple extension of the comment which was already here. But I am happy to discuss it further.

Done.

[janekmi]

@janekmi reviewed 1 of 1 files at r2, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @osalyk)