uname will not work correctly.Could you please post the "du" of your complete/finale Deep Clone?
That was the part which was disturbing. I think I have to go with Deep Clone.
In that case, I think the problem solved. Thanks.
This makes sense if you think about it. The commits for releng/13.0 were not fetched by your initial shallow clone because they were not referenced in the branch you shallow-cloned.
Where is this commit count in the uname output? Looking at my
uname -a, I don't see any. Just some hash that's included and the usual kernel build number that increases if you rebuild a kernel. And also that reproducible builds seems to be off by default now on 13.0, i.e. I get the build host, time and other information.Yes in a way it makes sense, but there still has to be a git way of switching a shallow clone to another branch? I have tried
git checkout as well as git switch and some git fetch command I found mentioned somewhere, but ultimately none of it worked. So I just did a fresh clone, which was done in a matter of seconds. git is fast, I give you that.% du -hs /home/git/repos/freebsd/src/
1,4G /home/git/repos/freebsd/src/
% zfs get used,logicalused,compression,compressratio sys/home/git
NAME PROPERTY VALUE SOURCE
sys/home/git used 1.40G -
sys/home/git logicalused 1.42G -
sys/home/git compression lz4 local
sys/home/git compressratio 1.02x - svnsync was way over 6G in size when I switched to git.
Here:
It involves adding the missing refs, so you can fetch the other branch you want. See for example this answer on SO: https://stackoverflow.com/a/17937889
See also my answer here:
Then I dont have that commit count. My output looks like:
FreeBSD [...]13.0-RELEASE FreeBSD 13.0-RELEASE #0 releng/13.0-ea31abc26 [...build info...]That looks cumbersome, to put it politelyIt involves adding the missing refs, so you can fetch the other branch you want. See for example this answer on SO: https://stackoverflow.com/a/17937889
See also my answer here:
Solved - How much HDD space do I need to Git clone main (deep vs. shallow)
How much HDD space do I need to 1. Git deep clone main. 2. Git shallow clone main.forums.FreeBSD.org
The required disk space is not exactly what I am concerned about. Required bandwidth on the other hand is, that's why I am keeping an on-premises mirror of the FreeBSD src git repo, that other machines use to update their /usr/src from. I am not working on those sources, they are merely there to pull updates and then rebuild the system (kernel/world) from it. There is hardly any point in ever going back, the only direction is forward (for example when 13.0-p1 will become available). So I just dont see the need to have that kind of history in something that is merely supposed to be a working copy.
git clone --no-single-branch --depth 1 ...?Full clone makes sense for you then. Git is very efficient at sending you only new changes when you do a fetch.
Basically what I have is a full clone minus the checkout of any branches, a bare mirror of the FreeBSD git repo which I update through
git remote update --prune that is then made available to my local network via git-daemon(1). All my other machines clone/update their /usr/src from this mirror, so it doesn't consume any external bandwidth at all. I just don't think each /usr/src on every machine really needs to be a full clone, when all I do is build from it.It's a clone done with the
-mirror option?So do a shallow clone on the leaf nodes.
Yes. When the switch to git came, I wanted to have an on-premises mirror just like I had with SVN, updated through
svnsync before and CVS using cvsup before that.Can you elaborate on what exactly those leaf nodes are? So far what I've used to clone /usr/src on my machines was
git clone -o freebsd --depth 1 -b releng/13.0 <url> <dir>, just that for the <url> part I dont use git.freebsd.org but my internal mirror instead.You're doing a shallow clone on your leaf nodes. What I'm proposing is do a fresh shallow clone when you want to switch branches.
rm -Rf <dir>
git clone -o freebsd --depth 1 -b releng/13.1 <url> <dir>That's what I ended up doing when switching from releng/12.2 to releng/13.0. Seems this is the quick and painless way of doing it then.
As this is my home network I would not exactly call it many machines, but in principle yes. My internet connection with ~7Mbps downstream is rather slow compared to any modern standards whatsoever and I'd hate to see bandwidth wasted, so it seems logical to do it that way.
