As i said, i am not advocating breaking up repositories. It would help
modulartiy, but is not worth the price.Submodules refer to versioned code so
changes to the submodule have no impact to the built code until until the
version referenced by the submodule is updated. I think that would alleviate
most of the concerns that you expressed.However, submodules have their own
issues I wouldn't ever attempt that again either. They are a maintenance
nightmare.Sent from my Galaxy
-------- Original message --------From: Nathan Hartman
<hartman.nat...@gmail.com> Date: 11/19/22 8:52 PM (GMT-06:00) To:
dev@nuttx.apache.org Subject: Re: New names of repositories On Sat, Nov 19,
2022 at 10:26 AM Gregory Nutt <spudan...@gmail.com> wrote:> On 11/19/2022 9:02
AM, Tomek CEDRO wrote:> > On Sat, Nov 19, 2022 at 3:53 PM Thiago Costa de Paiva
wrote:> >> From these, the valid argument is that it would be non-standard
(not a> top level name) related to Appache rules, which I think is indeed not>
acceptable. It would not imply any source code organization change nor>
submodules:> >>> >> git clone <appache url>/nuttx/core nuttx> >> git clone
<appache url>/nuttx/apps apps> > The use of submodules was also considered
before (I also proposed> > that) and turned out to be a bad idea :-) :-)>> At
one time, I was moving all major NuttX components into sub-modules> in order
to better modularize the system. I had arch/ and> Documentation/ moved into
their own repositories.>> But I got endless complaints from users. Everyone
hated this so I ended> up merging these back into nuttx/. That was a tough job
because of the> duplicated history in the nuttx/, arch/, and Documentation/>
repositories. I really wouldn't recommend going through that again.I strongly
recommend AGAINST splitting the NuttX repo into multiple repos.There are many
reasons for this but here is a big one: You'll lose atomicchangesets, which is
a very important part of version control. One of thebiggest problems with CVS,
which all modern version control systems aredesigned to avoid, is that changes
that touched multiple files were notatomic; each file had its own version
number, so you needed an out-of-bandway to keep track of which changes made up
a given changeset. Compare how asingle git commit hash number re-creates an
exact replica of the entiretree as it looked when that changeset was committed.
Splitting up the repois a step in the wrong direction.That said, nuttx and
nuttx-apps *should* remain separate for the followingreason: because the apps
are, by their nature, separate and optional. It isimportant that users will
always be able to supply their own appsrepository, while our apps repository
can be seen as demonstration,testing, a baseline against which to build a
custom apps, an example of howto build a custom apps, and even as documentation
in the form of sourcecode of how to write programs that run on NuttX. But
whatever happens,nuttx-apps should remain separate from nuttx.CheersNathan
