Re: [PATCH] build: get rid of the notion of a git library

> Junio C Hamano <gitster@pobox.com> writes:
>
>> And I do not see the reason why builtin/*.o should not depend on
>> each other.  It is not messed up at all.  They are meant to be
>> linked into a single binary---that is what being "built-in" is.
>>
>> A good way forward, the way it *SHOULD* be, is to slim the builtin/*.o
>> by moving parts that do not have to be in the single "git" binary
>> but are also usable in standalone binaries out of them.
>
> Actually, as long as these pieces are currently used by builtins,
> moving them (e.g. init_copy_notes_for_rewrite()) out of builtin/*.o
> will not make these parts not to be in the single "git" binary at
> all, so the above is grossly misstated.
>
>  - There may be pieces of usefully reusable code buried in
>    builtin/*.o;
>
>  - By definition, any code (piece of data or function definition) in
>    builtin/*.o cannot be used in standalone binaries, because all of
>    builtin/*.o expect to link with git.o and expect their cmd_foo()
>    getting called from main in it;
>
>  - By moving the useful reusable pieces ont of builtin/*.o and
>    adding them to libgit.a, these pieces become usable from
>    standalone binaries as well.

> And that is the reason why slimming builtin/*.o is the way it
> *SHOULD* be.
>
> Another thing to think about is looking at pieces of data and
> functions defined in each *.o files and moving things around within
> them.  For example, looking at the dependency chain I quoted earlier
> for sequencer.o to build upload-pack, which is about responding to
> "git fetch" on the sending side:
>
>         upload-pack.c   wants handle_revision_opt etc.
>         revision.c      provides handle_revision_opt
>                         wants name_decoration etc.
>         log-tree.c      provides name_decoration
>                         wants append_signoff
>         sequencer.c     provides append_signoff
>
> It is already crazy. There is no reason for the pack sender to be
> linking with the sequencer interpreter machinery. If the function
> definition (and possibly other ones) are split into separate source
> files (still in libgit.a), git-upload-pack binary does not have to
> pull in the whole sequencer.c at all.

> Coming back to the categorization Peff earlier made in the thread, I
> think I am OK with adding new two subdirectories to the root level,
> i.e.
>
>     builtin/    - the ones that define cmd_foo()

> We may also want to add another subdirectory to hold scripted
> Porcelains, but the primary topic of this thread is what to do about
> the C library, so it is orthogonal in that sense, but if we were to
> go in the "group things in subdirectories to slim the root level"
> direction, it may be worth considering doing so at the same time.