Boy... Gene... I've been at configuration management since my early days in the
late 90s and you are an *excellent* destructive tester for the 'standard'
configuration path! [In the neverending circle-of-life the thrill of testers'
victory is the agony of the developers defeat :-].
As for one thing you really should know: Git is *made made made* for
changes-that-don't-need-a-new-download. Git can get the latest any time you
like, but the pattern is really like this:
% git clone <url-here>
% git checkout -b <my-branch> origin/<source-branch>
% git log # see what's the latest thing and its function...
[ magic, edits and @#$@# happens here ]
% git diff --stat # summarized changes...
% git diff # full changes...
[ review the sanity you added to all the madness that came in ]
% git checkout ./packaging/deb/rules # a file restored back to original
state, readied for a commit
% git checkout ./autogen # a file restored back to original state,
readied for a commit ...
[ consider a retry for the build with your few changes ]
% git clean -d -f # remove only extra/temp files it detects
[none from the clone/repo]
[ magic, edits and final versions happen here ]
% git commit -a -m "wise and ingenious edits beyond question were... blah blah
blah"
[ changes are integrated into your branch, only on the local machine, btw ]
<time passes>
% git pull origin <orig-branch>
[ magic remains with new versions of crazy integrated. ]
% git log --oneline # to see all the things piled on yet again...
Any changes upstream then are copied in and *automatically* merged into place
over the top of yours, as long as you've committed them into place locally.
It's not fun if there's a conflict but them's the breaks if you and the remote
disagree at times. It allows you to keep up a highly modified build over time
if you like and, if you dare, you can use a 'rebase' to try and yank your
changes on top over and over.
In any case ... I didn't think I'd need to overhaul the build on the community
side, before. Still, it looks like more may need to be done to config the user
and to *at least* get the README in order! There's far too many cobwebs in
this setup and I *must* say you do seem to be working late at night also! I
may need a bit of time to catch up to your plans here. I'll get some more by
Sunday night probably.
-- CH
p.s.
Side notes: YES it does take a little bit of heavy-hit-NUDGING to convince
Out-The-Window-10/10 that you mean what you say. [sigh] Welding the hood shut
is their idea of fun.
And also, SSHFS is a pretty neat scheme and it's awesome that it can sit there
as a directory like any other. We're considering Ansible for some new work
here and SSH-only is great for interactive links, as opposed to
HTTPS-and-friends for every thing under the sun.
ASIDE ABOUT GIT
Git is a very very bizarre set of concepts, but its main function is simple:
maintain a winding 'tree' of 'undo' changes that were layered (across many
separate files) for every small shift or growth in a design that devs write.
As folks cooperate it attempts to tape-record *only* the deltas, the diffs, the
patches since the beginning of a project. Keeping them precisely a small as
the original writer wanted them to be. It then precisely knows how to un-lay
every brick and un-nail every joist and stud all the way back to a bare slab of
concrete. It does this so each intention for each change is pretty clear.
(They can be joined into a bigger change later on if needed).
NOTE: It does not maintain a 'video' in slow-mo of images of your files, like
Subversion does. It maintains a change-tree... so your own 'undo' chain can go
backward and (if you like) jump back forward ... but it's not required to be a
single timeline at all.
Why is this good to know things backwards? Because that's where variations or
mistakes are removed and better ways are tried: in reverse! Forward has
disagreements or variation but backward is always pretty clear.
Then you have a clear way to jump backwards a ways from one design (mistake) to
an earlier one (success). Even more you can yank better parts from one
developer branch over to another and 'replay them' on top of your own. It's
the undo key for programmers and *also* the way to tightly track any set of
local changes and re-add them on top of any solid version of anything anywhere
[i.e. rebase].
If reshuffling with a rebase (which can be messy) you can still track things by
stacking (but tracking) remote changes on your local changes and intermingle
them like a huge huge dagwood sandwich. Linear history be damned.. cuz coding
rarely goes linearly.
I have no apology for the bizarre set of names and terms and invisible states
it maintains. It is not easy to learn and an easier GUI version like "gitg" or
something is far better, if you can because it's very hard to visualize in the
first two years of use. Even so, things like local changes are good to make
and easy to stow away.
Restarting from a "reset --hard <orig-branch>" or a nicer "checkout -b mineno2
..." is always possible... but beware as some changes you make can be lost.
On 5/11/19 9:32 AM, Gene Heskett wrote:
Despite changing the name in the rules file, its hard coded to be
amanda-backup in the .dsc files. GRRRRRRR
Without fixing this, how the hell do we get rid of the older version
trash when updateing it? Seems like a heck of a good question to me. Do
we expect the users to wear out find in cleaning house file by file? I
think not.
This was amanda for 2 decades before some marketing type decreed a name
change. What was this person thinking?
Anyway, I blew that whole /home/amanda/amanda directory away an started
with becoming amanda, then a fresh git clone from github.
Then did exactly as said. It builds 5 packages. so I attempt to install
the amanda-backup-server since this machine is the server.
And its stuck, just like the previous attempt. No cpu used, no progress.
pstree does show but one process, but htop shows 5, with the highest pid
being from gnupg getting a no-permissions from /var/lib/amanda. An ls -l
of that dir is:
amanda@coyote:..$<mailto:amanda@coyote:..$> ls -l /var/lib/amanda
total 16
-rw-r--r-- 1 root root 34 May 11 09:59 amanda-release
-rw-r--r-- 1 backup backup 0 May 6 10:51 amandates
drwxr-xr-x 3 amandabackup disk 4096 May 11 11:01 example
drwxrwx--- 2 amandabackup disk 4096 May 11 09:59 gnutar-lists
drwxr-xr-x 2 amandabackup disk 4096 May 11 11:01 template.d
And that directory is owned by backup. And I'm lost, and I need to go
convince a winders 10 home edition that I am indeed boss. I can't even
get it online. Need a manpage for netfs.
Thanks Chris, but its Your turn.
Cheers, Gene Heskett
