Re: Extremely large change wave?

[Bob Proulx]

Ineiev wrote:
> Bob Proulx wrote:
> > I am trying to understand the large change wave that has been
> > committed in the last few days.
>
> Last quite a few months, in a sense. the date of 9fd58ef254ab is
> 2023-08-16.

On February 8th I made a commit fixing (partially) a critical bug.
Then through February 12th were more updates from you.  On Feb 12 was
the start of the very large wave of changes.

It was only a few days between those commits.  Saying it was months is
disingenuous at best.  Since it was days and are not months.

Also these sweeping changes basically pull the rug out from all others
who are trying to develop improvements for the platform.  Such as for
example Ayushman's feature improvement project for the FSF admins.
Such as my work to help with that, and with Corwin's work.  It's very
disruptive.  And it was done without any communication or coordination.

> > I made a commit on Feb 8 a894e1 and for example between then and now I
> > see many commits resulting in this large diff.
> >
> >     $ git diff a894e1..HEAD | diffstat | tail -n1
> >      386 files changed, 56532 insertions(+), 16683 deletions(-)
>
> Most of insertions are in two new data files,
> backend/external/cities15000{.txt,}, though.

That still leaves a lot of other disruptive changes.

> > That is a very large number of changes in a very short period of time!
> > This has only been in the last few days.
>
> I believe the most important change from Savannah admins' viewpoint
> wasn't in fact in the code. it was in how the frontend code is
> installed. previously, what was used were basically a Git clone
> (which wasn't actually supported); then I migrated to a regular
> installation.

That's a good example of a fundamental change which had no
communication nor coordination.  Though it is a change that I would
generally support.  But the result is that it leaves everyone else
lost at sea without any footing.

> I did it on Jan, 31, when migrating to frontend2, and I mentioned
> I was going to do that even earlier, when the fundraiser was
> running.

This detail must have been missed in the communications.

> > Just on the surface I see the following confusing things.
> >
> > * All of the .gitignore files have been deleted.  This causes a large
> >   amount of noise files to appear in the git status.  What's the plan
> >   for this?
>
> When I use a separate build tree, git status only shows these files
> [reordered]:
>
>   configure Makefile.in frontend/Makefile.in lib/Makefile.in
>   po/Makefile.in autotools/m4/Makefile.in
>   aclocal.m4 autom4te.cache/ autotools/install-sh autotools/missing
>   po/stamp-po po/savane.pot
>   po/ca.po~ po/de.po~ po/es.po~ po/fr.po~ po/he.po~ po/it.po~
>   po/ja.po~ po/pt_BR.po~ po/ru.po~ po/sv.po~
>
> I don't think the amount is really large and noisy.

Really?  I was seeing at least three pages of differences!  Requiring
the use of a pager to skip over files that should not be committed is
not a good thing.

> > * The local development router has been removed.  This was being used
> >   to run a local sandbox.
> >
> > * The script that was used to launch the local development router has
> >   been removed.
>
> To be precise, they weren't removed, they were replaced with scripts
> that take into account configure-time settings; moreover, an option
> to specify the path to the PHP executable was added (to say nothing
> of the standard --help and --version), a450ed9468.

It is good that we are communicating here.  Because otherwise no one
will know to look for this difference.  And now documentation can be
updated to make it possible for other people who are not you to clone
the repository and do work with it.

> > On the frontend servers the files being served had been in version
> > control at /opt/savannah/savane.  Those files are still there but the
> > version control has not been updated in a very long time.  Yet I think
> > as recently as last month when the fundraising banner was removed that
> > it had been updated.
>
> The files in /opt/savannah/savane are leftover from the times when
> frontend2 served as frontend.

That's quite the pitfall to leave behind.

> > I see that apparently on Jan 31 this has been changed to serve files
> > from /opt/savane/share/savane/frontend/php as what appears to be an
> > installed files area.  But I can't find any indication of where those
> > files are being installed from or from what version.
>
> mgt1:ChangeLog does include some records; general setup is now
> outlined in <https://savannah.gnu.org/maintenance/SavaneSetup/>.

I see that is a new file created on Feb 24.  I wrote my message on the
25th so of course being new I would not have known abou it.  It's only
the 27th today.  That's really a very new file!  Writing that as if it
has been there for a long time is again disingenuous at best.

> > When I look through the version control commit message history I see a
> > lot of very confusing things that indicate to me that the timeline was
> > committed and then reset and then re-committed differently.
>
> I often modify and reorder commits before they enter the master branch,
> typically because
>
> * a commit may need fixing, or it may be incomplete; committing
>   every fix at HEAD would make the history more complicated than
>   necessary; or
>
> * some commits seem ready for master, i.e. it doesn't look like
>   they'll need fixing; then I move them closer to master (or just
>   push them into master).
>
> Git preserves the original dates, as far as I can tell.

All that you say here is perfectly true and perfectly false at the
same time.  The files you work with in your local sandbox may have
been there for years and no one else will have been able to see them,
to view them, to comment upon them, or even to know if they are
happening at all.  Those private dates do not matter.  The only dates
that matter are the dates when they become visible to us.  Those are
the dates to which I am referring.

The important date for us is that on February 12th a large change wave
was committed that completely disrupted everyone else that was trying
to work on the project.  And then we see in the above that even your
later documentation for it did not appear until Feb 24th, which is
almost "just now".

I will read through your new INSTALL file and will work through your
new process.  I won't have any more comments until I have a chance to
do this.

Let's not lose sight that our purpose is, for the time we are
privileged to be maintainers of it, to take care of this grand project
that to which we have been given access.  The history of Savannah
maintainers and contributors has been a long one with many maintainers
before us and hopefully many more maintainers after us.  We are here
to facility the maintenance and improvement of this public good.  We
are here to mentor newcomer developers.  To develop a strong team of
administrators such that the GNU Free Software Forge will grow and
prosper into the future.

Bob