On Thu, 15 Aug 2024 at 15:11, Patrick Steinhardt <ps@xxxxxx> wrote:
>
> On Wed, Aug 14, 2024 at 05:33:17PM +0530, Chandra Pratap wrote:
> > In the current testing setup, block operations are left unexercised
> > for obj blocks. Add a test that exercises these operations for obj
> > blocks.
>
> Same remarks here as for the preceding commit.
>
> > @@ -186,9 +186,88 @@ static void t_log_block_read_write(void)
> > reftable_record_release(&recs[i]);
> > }
> >
> > +static void t_obj_block_read_write(void)
> > +{
> > + const int header_off = 21;
> > + struct reftable_record recs[30];
> > + const size_t N = ARRAY_SIZE(recs);
> > + const size_t block_size = 1024;
> > + struct reftable_block block = { 0 };
> > + struct block_writer bw = {
> > + .last_key = STRBUF_INIT,
> > + };
> > + struct reftable_record rec = {
> > + .type = BLOCK_TYPE_OBJ,
> > + };
> > + size_t i = 0;
> > + int n;
> > + struct block_reader br = { 0 };
> > + struct block_iter it = BLOCK_ITER_INIT;
> > + struct strbuf want = STRBUF_INIT;
> > +
> > + REFTABLE_CALLOC_ARRAY(block.data, block_size);
> > + block.len = block_size;
> > + block.source = malloc_block_source();
> > + block_writer_init(&bw, BLOCK_TYPE_OBJ, block.data, block_size,
> > + header_off, hash_size(GIT_SHA1_FORMAT_ID));
> > +
> > + for (i = 0; i < N; i++) {
> > + uint8_t *bytes = reftable_malloc(sizeof(uint8_t[5]));
> > + memcpy(bytes, (uint8_t[]){i, i+1, i+2, i+3, i+5}, sizeof(uint8_t[5]));
>
> From the top of my head I'm not sure whether we use inline-array
> declarations like this anywhere. I'd rather just make it a separate
> variable, which also allows us to get rid of the magic 5 via
> `ARRAY_SIZE()`.
We _do_ use inline array declarations like this, here's an example from
t/unit-tests/t-prio-queue.c:
TEST(TEST_INPUT(((int []){ STACK, 1, 2, 3, 4, 5, 6, REVERSE, DUMP }),
((int []){ 1, 2, 3, 4, 5, 6 })), "prio-queue works when LIFO
stack is reversed");
I did implement bytes[] as a local variable array when I first worked
on this patch but that turned out to be tricky due to variable scoping
and pointer semantics, so I ultimately settled on this approach.
