Change iterators API about pmemstream

[igchor]

pmemstream_entry_iterator_get could actually just return pmemstream_entry by value. Its precondition is pmemstream_entry_iterator_is_valid() == 0 anyway, right?

[karczex]

It's right as long as you cannot remove region when region or entry iterator points to it. IMO, we may do it this way.

[igchor]

Yes, the idea is that iterator will block removal. If it would not, we would not have any way of detecting this anyway I think.

[KFilipek]

What's the correct behavior for:

struct pmemstream_region pmemstream_region_iterator_get(struct pmemstream_region_iterator *iterator, struct pmemstream_region *region);

when:

void invalid_region_test(char *path)
{
	pmemstream_test_env env = pmemstream_test_make_default(path);

	const uint64_t invalid_offset = ALIGN_DOWN(UINT64_MAX, sizeof(span_bytes));
	struct pmemstream_region_iterator *riter = NULL;
	struct pmemstream_region invalid_region = {.offset = invalid_offset};

	int ret = pmemstream_region_iterator_new(&riter, env.stream);
	UT_ASSERTeq(ret, 0);
	UT_ASSERTne(riter, NULL);

	pmemstream_region_iterator_seek_first(riter);

	invalid_region = pmemstream_region_iterator_get(riter);
	UT_ASSERTeq(invalid_region.offset, invalid_offset); // HERE

	pmemstream_region_iterator_delete(&riter);
	pmemstream_test_teardown(env);
}

[lukaszstolarczuk]

the above question is resolved now - an invalid region is just returned from get and such region can always be checked by pmemstream_entry_iterator_is_valid function.

// region part is done in #186

[igchor]

@lukaszstolarczuk why this is closed? entry iterator API is not done yet, right?

[lukaszstolarczuk]

yeah, for some reason GH closed this (the commit has a reference to this issue).