On Sat, 10 Aug 2024 at 00:27, Junio C Hamano <gitster@xxxxxxxxx> wrote:
>
> Chandra Pratap <chandrapratap3519@xxxxxxxxx> writes:
>
> > free_names() as defined by reftable/basics.{c,h} frees a NULL
> > terminated array of malloced strings along with the array itself.
> > Use this function instead of a for loop to free such an array.
>
> Going back to [1/4], the headers included in this test looked like this:
>
> -#include "system.h"
> -
> -#include "basics.h"
> -#include "block.h"
> -#include "blocksource.h"
> -#include "reader.h"
> -#include "record.h"
> -#include "test_framework.h"
> -#include "reftable-tests.h"
> -#include "reftable-writer.h"
> +#include "test-lib.h"
> +#include "reftable/reader.h"
> +#include "reftable/blocksource.h"
> +#include "reftable/reftable-error.h"
> +#include "reftable/reftable-writer.h"
>
> I found this part a bit curious, perhaps because I was not involved
> in either reftable/ or unit-tests/ development. So I may be asking
> a stupid question, but is it intended that some headers like
> "block.h" and "record.h" are no longer included?
>
> It is understandable that inclusion of "test-lib.h" is new (and
> needs to be there to work as part of t/unit-tests/), and the leading
> directory name "reftable/" added to header files are also justified,
> of course. But if you depend on "basics.h" and do not include it,
> that does not sound like the most hygenic thing to do, at least to
> me.
I think 'basics.{c,h}' in reftable/ is equivalent to 'stdio.h' in a generic
C program, it holds fundamental functionalities to be used by other
reftable structures and hence, is always (implicitly or explicitly)
#included in almost all of the files in reftable/.
This test is supposed to focus on reftable's read-write functionalities
so it makes sense to explicitly #include only those headers that
are directly responsible for those functionalities, namely 'reader.h',
'blocksource.h' and 'reftable-writer.h'. 'reftable-error.h' is thrown in
there as well because some tests need to explicitly mention the
various error codes and it doesn't make sense to rely on it being
#included by others.
> The code changes themselves look good; I can see that the
> implementation of free_names() in reftable/basics.c safely replaces
> these loops. There is a slight behaviour difference that names[]
> that was fed to reftable_iterator_seek_ref() earlier goes away
> before the iterator is destroyed, but _seek_ref() does not retain
> the names[0] argument in the iterator object, so that is OK.
>
> Thanks.
