On 28.01.22 02:15, gene heskett wrote:
On Thursday, January 27, 2022 7:17:10 PM EST Steffen Möller wrote:
On 28.01.22 00:31, gene heskett wrote:
On Thursday, January 27, 2022 2:22:43 PM EST Steffen Möller wrote:
On 27.01.22 20:03, gene heskett wrote:
On Thursday, January 27, 2022 9:54:54 AM EST Steffen Möller wrote:
Fresh start!
Dear Gene,
I would like to catch the problem before you start building and
also
exclude the possibility that somehow the code base of yours is
affected by your previous checkout - just because I cannot inspect
your machine from here. Once that was successful, yes, then this
can be optimized.
First thing is that the system needs to be truly updated, nothing
half-ish.
Remember Steffen, that this sd card was A, new, and b, written with
dd
using 2021-10-30-raspios-bullseye-armhf-full.img,
sd card then put in the pi and booted, after I had fixed the no
network problem by filling in the defaults for a static network by
putting my hosts file over the default, editing /etc/hostname to be
the same as the buster install it would replace, and filling in and
uncommenting the bottom of its /etc/dhcpcd.conf file to match the
buster net config. It was then booted, I assigned the country,
keyboard and other first boot things in raspi-config. rebooted, at
which point it ran the update/ upgrade stuff bringing it up to date
by upgrading 129 pkgs then. The apt update/apt upgrade -y has been
done several more times, and just now replaced 5 python pkgs.
sudo apt dist-upgrade just returned:
=============================
pi@rpi4:/media/pi/workspace $ sudo apt dist-upgrade -y
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
Calculating upgrade... Done
The following packages were automatically installed and are no
longer
required:
dctrl-tools dkms libfuse2 libxtables-dev
raspberrypi-kernel-headers
Use 'sudo apt autoremove' to remove them.
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
=============================
So it is, ANAICT, a fully uptodate bullseye install.
In the middle of that, I mounted the drive I had built a
5.16.2-rt19
kernel on and installed that, so a uname -a now returns:
===========================
Linux rpi4 5.16.2-rt19-rt19-v7l+ #3 SMP PREEMPT_RT Tue Jan 25
01:14:16
EST 2022 armv7l GNU/Linux
===========================
Almost as bleeding edge as has been announced on the linux-rt list.
I have since ran the first of my scripts to build master, then
scanned
the output for missing dependency's, installing those I spot in the
back trace.
I have not adjusted anything in the config, or debian directories
of
this git clone which is of the raspberry/linux. I have been doing
this build since jessie days so I'm not new to this. At one point
pi
stuff was being built on an odroid C2 at the buildbot, but crashed
several times a week and has been replaced with an rpi4 I believe,
same as this one.
I faintly recall having to do something for buster but at 87 yo I
do
not recall what it was I had to do back then. And I have not noted
the build your own recipe in our wiki as having been updated since
wheezy, so it is, shall we say, a bit long in the tooth in 2022. :)
And the next missing dependency is "convert", and its a showstopper
for configure.
I was not aware of "convert". You have done everything just fine.
Please do
sudo apt update
that showed 5 pkgs could be upgraded which I did.
sudo apt -u dist-upgrade
see above
Anything surprising/weird/many packages listed? Then please tell
me
or
continue with "yes".
Wherever you have the disk space please then do a
I have the space, its a 240gig SSD
git clone https://github.com/LinuxCNC/linuxcnc.git
bullseye-linuxcnc
Will take an hour or more, my net connection is leisurely.
cd new-dir-name
python3 --version # should be more recent than 3.7
pi@rpi4:/media/pi/workspace $ python3 --version
Python 3.9.2
Please ping me again once you got to this stage.
And configure still bails out:
checking for convert... none
configure: error: no convert, documentation cannot be built
You are two steps ahead :) I presume you ran
debian/configure
but just do it again, please, so I know what was done. The "uspace"
argument is the default, so just run as shown above.
We now have the debian/control file. This debian/control file
declares
the packages that are required to run the package, but especially
also
the packages that are required to build the package.
Now run
dpkg-buildpackage
and it will check that debian/control file to see if all packages
are
indeed installed that need to be installed to build the packages.
When you invoke this now then it will fail (so I hope) because the
package "imagemagick" is missing which then also provides
/usr/bin/convert. There are likely other packages, too, that
configure
would identify as missing if it was not already halting after
checking
for "convert". So, dpkg-buildpacakge will list all the packages that
are missing. Please install those and then run dpkg-buildpackage
again.
LinuxCNC has a bit of a problem with interrupted builds. There is a
chance that dpkg-buildpackage fails because files existing that are
not in the original tarball. But we will get to that - now it is me
who is ahead of himself :)
Best,
Steffen
I installed imagemagick and all but one mesa util apt couldn't find,
did a make clean
dpkg-buildpackage invokes a cleaning itself. This should not be
required. In theory.
and reran configure,
Hummm, my "maketoruntests.sh" may be running the wrong configuration:
=======================
cd /media/pi/workspace/bullseye-linuxcnc/
git checkout master
cd ./src
./autogen.sh
./configure --enable-build-documentation
sudo make clean
make -j4
sudo make setuid
source ../scripts/rip-environment
cd ..
cd ./src
runtests
=======================
I'll modify that to run debian/configure --enable-build-documentation
Please imagine that I sit here on the other side of the globe and try to
understand where the error messages are coming from. If you now show me
anything else than the commands I asked you to execute and their outputs
then this is confusing me.
Which configure? The ./configure ? Maybe this works. Maybe not.
Configure takes options and for proper debian packages these need to be
set properly :) It is not magic, how it is done you can inspect in the
file debian/rules, which is derived from debian/rules.in by a call to
debian/configure.
Please do a
git reset --hard
That seem to have put me on 2.8, not master?
pi@rpi4:/media/pi/workspace/bullseye-linuxcnc $ git reset --hard
HEAD is now at 65b209390 Merge remote-tracking branch 'origin/2.8'
which should bring you back to pristine master branch. Please also do a
git status
pi@rpi4:/media/pi/workspace/bullseye-linuxcnc $ git status
On branch master
Your branch is up to date with 'origin/master'.
nothing to commit, working tree clean
Which is puzzling, shouldn't both report pretty close to the same thing?
The "reset --hard" cleaned everything. That is perfect.
and remove any files that are listed as not tracked (there should not
be any, just to be sure).
Then start over with debian/configure (the debian one, not the one in
your working directory, which should now no longer exist since it was
generated). No need to directly invoke any configure or any make from
here on. The file debian/rules is a Makefile itself and knows how to
invoke the real Makefile. And that debian/rules file is invoked via
dpkg-buildpackage.
So, after debian/configure please run dpkg-buildpackage again and this
time it should complete. If not, well, then something is special on
your RPi platform. We'll get there, eventually.
Best,
Steffen
My script snippet
As I said. I am on the other side of the globe - my problem solving
implies that you run my commands. You can run more to give more
information but anything that writes to anything in that directory is a
no-no.
What you are doing is that you have a configuration problem but are not
running the configuration that is performed in debian/rules, which is
executed by dpkg-buildpackage, which I had asked to run, and only that.
I suggest you try again with the commands I suggested. In the meantime,
let me build the LinuxCNC packages myself for ARM64 on bullseye and I
place them somewhere for you to download. I should have these by next week.
Best,
Steffen
ran as usual until it ran into boost::python problems,
then 137 of the runtests failed, most blaming boost::python::this or
that.
Cheers Steffen, Gene Heskett.
_______________________________________________
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers
