gc_alloc() - should collected=1 after gc_collect() is executed?

[adritium]

micropython/py/gc.c

Lines 452 to 458 in 1b7487e

	#if MICROPY_GC_ALLOC_THRESHOLD
	if (!collected && MP_STATE_MEM(gc_alloc_amount) >= MP_STATE_MEM(gc_alloc_threshold)) {
	GC_EXIT();
	gc_collect();
	GC_ENTER();
	}
	#endif

If a gc_collect() is performed here then why isn't collected set to 1 just like it is in the next code block?

micropython/py/gc.c

Lines 471 to 479 in 1b7487e

	GC_EXIT();
	// nothing found!
	if (collected) {
	return NULL;
	}
	DEBUG_printf("gc_alloc(" UINT_FMT "): no free mem, triggering GC\n", n_bytes);
	gc_collect();
	collected = 1;
	GC_ENTER();

Under existing code, it seems you would

1. Perform a gc_collect() (from L455)
2. Attempt to locate a large enough block
3. If large enough block not found in step 2, perform another gc_collect() which would be 100% pointless because no allocation operation was performed between step 2 and step 3 so it's guaranteed the memory layout hasn't changed.
4. Set collected = 1
5. Look again for a large enough block as in step 2 - since it failed the first time, it will fail again because memory layout has not change since the first time gc_collect() was called on L455
6. If large enough block not found, return NULL

By adding collected = 1 at L455, we'd be avoiding an extra gc_collect() + search.

[dpgeorge]

By adding collected = 1 at L455, we'd be avoiding an extra gc_collect() + search.

Yes, that looks to be a correct conclusion.

[dpgeorge]

Fixed by 6bd7874