[LTP] [PATCH v3 1/2] tst_memutils.c: Add tst_print_meminfo function

[Petr Vorel]

> Li Wang <liwang@redhat.com> wrote:

> >> > +++ b/lib/safe_macros.c

> >> We don't want to add anything to the legacy API (otherwise it would go to
> >> lib/safe_file_ops.c), please add this to lib/tst_safe_macros.c.

> >> BTW I'm slightly confused, what would be the best place for this,
> >> lib/tst_safe_macros.c is being used nowadays for everything. But there is
> >> also



> >> include/tst_safe_file_ops.h, which does not have C file
> >> (lib/tst_safe_file_ops.c) because it internally use lib/tst_safe_macros.c.


> > No, basically it does not use the lib/tst_safe_macros.c.

> > From what I understand, 'tst_safe_file_ops.h' is just a header for
> > collecting
> > all the file operations for Linux, it doesn't touch other subcomponent ops.

> > 'tst_safe_file_ops.h' defines macros for all functions in
> > 'safe_file_ops_fn.h'
> > and archived them in 'safe_file_ops.c' lib.

> > Something like this combination:

> > tst_safe_file_ops.h:
> >     safe_file_ops_fn.h
> >     safe_file_ops.c


> The reason to split them into two headers is for backward compatibility.

> Technically we should declare 'safe_file_ops_fn.h' functions in one header
> 'tst_safe_file_ops.h' but you know we have 'old_safe_file_ops.h'.

> tst_safe_file_ops.h:
>     safe_file_ops_fn.h
>     safe_file_ops.c

> old_safe_file_ops.h:
>     safe_file_ops_fn.h
>     safe_file_ops.c

> In the future, the ideal cleanup direction is just to have:

> tst_safe_file_ops.h
> tst_safe_file_ops.c

Yep, the problem that Cyril originally implemented new API via legacy to
support both APIs (good and needed solution) and now we add only into new API
(also good solution) + that approach specific (and thus small) C sources
initially and later more generic (and longer) makes library a bit inconsistent.
I know, it will be solved once we get rid of the legacy API, but that will take
time.

Kind regards,
Petr