On Friday, January 28, 2022 5:34:34 AM EST Steffen Möller wrote: > 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.
I am doing both, and I've modified my script to run that config with a make clean between them, what I am reporting to you from here on will be from your commands, I will make my scripts compatible with yours except I'm actually building installable debs when my stuff works. I think I'm also using the buildpackage later but I build first for runinplace so I can run the test suite insitu, if that is 100% successful then I build the debs. runtests takes at least half an hour but its checking is worth it IMO. > 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. Not on arm64, Steffen but on armhf (v7l) please. The deeper stack frame on arm64 has a several microsecond negative effect on irq latency where lower is better. And we need every microsecond we can extract. armhf on an older 4.19 realtime preempt kernel I built, manages around 12 microseconds, not great, but adequate where some intel stuff can do 4 u- secs. Because this bullseye has not yet allowed linuxcnc to be built, I haven't a clue what my 5.16-2-rt19 will do yet, but I expect it will beat the 4.19 version by a couple microseconds where lower is better. arm64's latency I expect to be more than 20 microseconds which may make the system stutter and stall stepper motors. That can wreck parts AND break expensive tooling. > Best, > Steffen This result however was with my script, which did get as far as runtests, with 137 out of 250 or so failing. > > 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 > [email protected] > https://lists.sourceforge.net/lists/listinfo/emc-developers Cheers, Gene Heskett. -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author, 1940) If we desire respect for the law, we must first make the law respectable. - Louis D. Brandeis Genes Web page <http://geneslinuxbox.net:6309/gene> _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
