Re: [ptxdist] Yocto
Hi, On Fri, Sep 15, 2017 at 11:02:14AM +, Alm, Michael wrote: > when was the -j option added to ptxdist? Which version? It was first introduced in ptxdist-2015.06.0 but, if I remember correctly there were still some parallel building issues to fix at the time. It works quite well for me these days, but we find some remaining issues occasionally. In most cases, just restarting ptxdist is sufficient. Michael -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0| Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917- | ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hi, when was the -j option added to ptxdist? Which version? Regards, Michael -Ursprüngliche Nachricht- Von: ptxdist [mailto:ptxdist-boun...@pengutronix.de] Im Auftrag von Michael Olbrich Gesendet: Freitag, 15. September 2017 08:47 An: ptxdist@pengutronix.de Betreff: Re: [ptxdist] Yocto Hi, On Fri, Sep 15, 2017 at 12:44:13AM +0200, Tim Sander wrote: > Am Mittwoch, 13. September 2017, 17:38:24 CEST schrieb Michael Olbrich: > > Btw. don't use '-jeX -jiX' use '-j' instead. That uses a global > > jobserver shared by all packages. > Well i missed out this one. I know that this option didn't exist. But > i didn't realize that its now the recommended way. Its much better > now, just with a few passages where the cores are waiting for install io. > > Probably you should update the default output of ptxdist. Because at > least > 2017.07.0 doesn't list -j as an option. So now i know why the -j > options slipped my attention :-). I added it to the documentation but not the help text :-/. Fixed now. > Besides what is -je good for now? Nothing really. I left it there at first because I wasn't sure if there were any problems with -j. Now it's just for backwards compatibility. Regards, Michael -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0| Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917- | ___ ptxdist mailing list ptxdist@pengutronix.de ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hi, On Fri, Sep 15, 2017 at 12:44:13AM +0200, Tim Sander wrote: > Am Mittwoch, 13. September 2017, 17:38:24 CEST schrieb Michael Olbrich: > > Btw. don't use '-jeX -jiX' use '-j' instead. That uses a global jobserver > > shared by all packages. > Well i missed out this one. I know that this option didn't exist. But i > didn't > realize that its now the recommended way. Its much better now, just with a > few > passages where the cores are waiting for install io. > > Probably you should update the default output of ptxdist. Because at least > 2017.07.0 doesn't list -j as an option. So now i know why the -j options > slipped my attention :-). I added it to the documentation but not the help text :-/. Fixed now. > Besides what is -je good for now? Nothing really. I left it there at first because I wasn't sure if there were any problems with -j. Now it's just for backwards compatibility. Regards, Michael -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0| Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917- | ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hi Micheal Am Mittwoch, 13. September 2017, 17:38:24 CEST schrieb Michael Olbrich: > On Tue, Sep 12, 2017 at 08:42:26PM +0200, Tim Sander wrote: > > On Sonntag, 10. September 2017 21:16:52 CEST Robert Schwebel wrote: > > > On Sat, Sep 09, 2017 at 08:36:13PM +0200, Tim Sander wrote: > > > > As for the speed i would say that one of my biggest gripes with > > > > ptxdist is > > > > that -je16 -ji16 builds fail for me. And with that many virtual cores > > > > available its realy a pitty :-(. > > > > > > Do you have logfiles for the fails? We are doing a lot of test builds > > > here, and if something breaks, we would like to know. > > > > Strange, i think if i just do a build of DistroKit i can observe that make > > just stops. I can restart it so a > > for i in `seq 1 10`; do ptxdist -je16 -ji16 go; done; > > Works faster than a normal build, so the failure is that not all packages > > are getting build? > > It's probably a dependency problem exposed by the -jeX. If you see > something like this again, a logfile would be great. Stuff like that often > depends on the hardware so it's difficult to reproduce. > I've not seen Problems like this for some time. > > Btw. don't use '-jeX -jiX' use '-j' instead. That uses a global jobserver > shared by all packages. Well i missed out this one. I know that this option didn't exist. But i didn't realize that its now the recommended way. Its much better now, just with a few passages where the cores are waiting for install io. Probably you should update the default output of ptxdist. Because at least 2017.07.0 doesn't list -j as an option. So now i know why the -j options slipped my attention :-). Besides what is -je good for now? Best regards Tim ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
On Tue, Sep 12, 2017 at 08:42:26PM +0200, Tim Sander wrote: > On Sonntag, 10. September 2017 21:16:52 CEST Robert Schwebel wrote: > > On Sat, Sep 09, 2017 at 08:36:13PM +0200, Tim Sander wrote: > > > As for the speed i would say that one of my biggest gripes with ptxdist is > > > that -je16 -ji16 builds fail for me. And with that many virtual cores > > > available its realy a pitty :-(. > > > > Do you have logfiles for the fails? We are doing a lot of test builds > > here, and if something breaks, we would like to know. > Strange, i think if i just do a build of DistroKit i can observe that make > just stops. I can restart it so a > for i in `seq 1 10`; do ptxdist -je16 -ji16 go; done; > Works faster than a normal build, so the failure is that not all packages are > getting build? It's probably a dependency problem exposed by the -jeX. If you see something like this again, a logfile would be great. Stuff like that often depends on the hardware so it's difficult to reproduce. I've not seen Problems like this for some time. Btw. don't use '-jeX -jiX' use '-j' instead. That uses a global jobserver shared by all packages. Michael -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0| Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917- | ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
On Tue, Sep 12, 2017 at 11:33:42PM +0200, Tim Sander wrote: > Am Dienstag, 12. September 2017, 22:12:15 CEST schrieb Robert Schwebel: > > On Tue, Sep 12, 2017 at 08:42:26PM +0200, Tim Sander wrote: > > > On Sonntag, 10. September 2017 21:16:52 CEST Robert Schwebel wrote: > > > > On Sat, Sep 09, 2017 at 08:36:13PM +0200, Tim Sander wrote: > > > > > I really would like to use ptxdist for my next private project where i > > > > > would like to try ROS. > > > > > > > > ROS1 or ROS2? > > > > > > I would like to have something not to dated. The website lists Lunar or > > > Kinetic Kame as versions... > > > > That's all ROS1. ROS2 is a complete redesign. > > You are right, but i am wetting my toes... and the ros.org website does not > mention ros2. Do you think ros2 is already usable? It's not about being ros2 a finished thing that you just can use, it's about thinking about what might be relevant in the future. Technology wise, ros2 solves some issues which have just "happened" with ros1 without being designed. However, any ros package set for ptxdist would be welcome :-) > > > > > As far as i see, its already packaged for OpenEmbedded or Angstrom. > > > > > > > > I only know of old ROS packages made by the BMW folks, but this is > > > > neither current versions of ROS, nor ROS2. Do you know of newer code? > > > > > > This seems to be the repository for kinetic kame: > > > https://github.com/bulwahn/meta-ros/tree/kinetic-experimental > > > > Lunar Loggerhead is the current release, right? > > Yes and kinetic is kind of a stable release, whatever that means. Basically, like with all Linux things, you should not care about "stable" at the beginning of a development. "Stable" is something you can think of when your development is finished and you care about using it for some real work. But as long as there are not even packages, start with "top-of-tree". > > > > We started with ptxdist packaging for ROS2 two techweeks ago, but it is > > > > a slow process if you have only one week per year. If someone would be > > > > interested commercially, it would probably be possible to get a solution > > > > faster, bug that didn't happen so far. > > > > > > Well its my private tinkering project... > > > Do you have something i can build upon. I think one hurdle would be the > > > custom build system? > > > > Of course, that's the main issue. Sorry, nothing finished yet... > > I am happy with anything borken. Its better than nothing... I'll look around if we have something that could be in a state to be shared. rsc -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0| Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917- | ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Am Dienstag, 12. September 2017, 22:12:15 CEST schrieb Robert Schwebel: > On Tue, Sep 12, 2017 at 08:42:26PM +0200, Tim Sander wrote: > > On Sonntag, 10. September 2017 21:16:52 CEST Robert Schwebel wrote: > > > On Sat, Sep 09, 2017 at 08:36:13PM +0200, Tim Sander wrote: > > > > I really would like to use ptxdist for my next private project where i > > > > would like to try ROS. > > > > > > ROS1 or ROS2? > > > > I would like to have something not to dated. The website lists Lunar or > > Kinetic Kame as versions... > > That's all ROS1. ROS2 is a complete redesign. You are right, but i am wetting my toes... and the ros.org website does not mention ros2. Do you think ros2 is already usable? > > > > As far as i see, its already packaged for OpenEmbedded or Angstrom. > > > > > > I only know of old ROS packages made by the BMW folks, but this is > > > neither current versions of ROS, nor ROS2. Do you know of newer code? > > > > This seems to be the repository for kinetic kame: > > https://github.com/bulwahn/meta-ros/tree/kinetic-experimental > > Lunar Loggerhead is the current release, right? Yes and kinetic is kind of a stable release, whatever that means. > > > > But as it has its own build tools and lots of python module > > > > dependencies it might be to much to get all this stuff build with > > > > ptxdist. > > > > > > We started with ptxdist packaging for ROS2 two techweeks ago, but it is > > > a slow process if you have only one week per year. If someone would be > > > interested commercially, it would probably be possible to get a solution > > > faster, bug that didn't happen so far. > > > > Well its my private tinkering project... > > Do you have something i can build upon. I think one hurdle would be the > > custom build system? > > Of course, that's the main issue. Sorry, nothing finished yet... I am happy with anything borken. Its better than nothing... > > > > As for the speed i would say that one of my biggest gripes with > > > > ptxdist is > > > > that -je16 -ji16 builds fail for me. And with that many virtual cores > > > > available its realy a pitty :-(. > > > > > > Do you have logfiles for the fails? We are doing a lot of test builds > > > here, and if something breaks, we would like to know. > > > > Strange, i think if i just do a build of DistroKit i can observe that make > > just stops. I can restart it so a > > for i in `seq 1 10`; do ptxdist -je16 -ji16 go; done; > > Works faster than a normal build, so the failure is that not all packages > > are getting build? > > Hmm, we'll have a look. That would be really nice :-) Tim ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
On Tue, Sep 12, 2017 at 08:42:26PM +0200, Tim Sander wrote: > On Sonntag, 10. September 2017 21:16:52 CEST Robert Schwebel wrote: > > On Sat, Sep 09, 2017 at 08:36:13PM +0200, Tim Sander wrote: > > > I really would like to use ptxdist for my next private project where i > > > would like to try ROS. > > > > ROS1 or ROS2? > > I would like to have something not to dated. The website lists Lunar or > Kinetic Kame as versions... That's all ROS1. ROS2 is a complete redesign. > > > As far as i see, its already packaged for OpenEmbedded or Angstrom. > > > > I only know of old ROS packages made by the BMW folks, but this is > > neither current versions of ROS, nor ROS2. Do you know of newer code? > > This seems to be the repository for kinetic kame: > https://github.com/bulwahn/meta-ros/tree/kinetic-experimental Lunar Loggerhead is the current release, right? > > > But as it has its own build tools and lots of python module > > > dependencies it might be to much to get all this stuff build with > > > ptxdist. > > > > We started with ptxdist packaging for ROS2 two techweeks ago, but it is > > a slow process if you have only one week per year. If someone would be > > interested commercially, it would probably be possible to get a solution > > faster, bug that didn't happen so far. > > Well its my private tinkering project... > Do you have something i can build upon. I think one hurdle would be the > custom > build system? Of course, that's the main issue. Sorry, nothing finished yet... > > > As for the speed i would say that one of my biggest gripes with ptxdist is > > > that -je16 -ji16 builds fail for me. And with that many virtual cores > > > available its realy a pitty :-(. > > > > Do you have logfiles for the fails? We are doing a lot of test builds > > here, and if something breaks, we would like to know. > > Strange, i think if i just do a build of DistroKit i can observe that make > just stops. I can restart it so a > for i in `seq 1 10`; do ptxdist -je16 -ji16 go; done; > Works faster than a normal build, so the failure is that not all packages are > getting build? Hmm, we'll have a look. rsc -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0| Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917- | ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
On Sonntag, 10. September 2017 21:16:52 CEST Robert Schwebel wrote: > On Sat, Sep 09, 2017 at 08:36:13PM +0200, Tim Sander wrote: > > I really would like to use ptxdist for my next private project where i > > would like to try ROS. > > ROS1 or ROS2? I would like to have something not to dated. The website lists Lunar or Kinetic Kame as versions... > > As far as i see, its already packaged for OpenEmbedded or Angstrom. > > I only know of old ROS packages made by the BMW folks, but this is > neither current versions of ROS, nor ROS2. Do you know of newer code? This seems to be the repository for kinetic kame: https://github.com/bulwahn/meta-ros/tree/kinetic-experimental > > But as it has its own build tools and lots of python module > > dependencies it might be to much to get all this stuff build with > > ptxdist. > > We started with ptxdist packaging for ROS2 two techweeks ago, but it is > a slow process if you have only one week per year. If someone would be > interested commercially, it would probably be possible to get a solution > faster, bug that didn't happen so far. Well its my private tinkering project... Do you have something i can build upon. I think one hurdle would be the custom build system? > > As for the speed i would say that one of my biggest gripes with ptxdist is > > that -je16 -ji16 builds fail for me. And with that many virtual cores > > available its realy a pitty :-(. > > Do you have logfiles for the fails? We are doing a lot of test builds > here, and if something breaks, we would like to know. Strange, i think if i just do a build of DistroKit i can observe that make just stops. I can restart it so a for i in `seq 1 10`; do ptxdist -je16 -ji16 go; done; Works faster than a normal build, so the failure is that not all packages are getting build? Best regards Tim ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hi, > Am Montag, 11. September 2017, 07:14:00 schrieb Schenk, Gavin: > > * No Tab-completition as PTXDIST offers it. 5 Stages, ~130packages and > > 100.000 commands. I think a very usefull feature in PTXDIST! > > How does this work? I use zsh and have no tab completion for ptxdist, > but miss it every now and then. :-/ > In house workshops with e.g. Michael Olbrich are the solution for many problems ;-). I am not familiar with zsh, here is the solution for bash that works for me: 1.) apt-get install bash-completion 2.) Create a file ~/.bash_completion with content export PATH=/opt/ptxdist/bin:$PATH . /opt/ptxdist/lib/ptxdist-2017.07.0/scripts/bash_completion 3.) If you use a symbolic link in your project to the ptxdist binary that is not called ptxdist ( I use only 'p' ) you have to add a line in ~/.bash_completion complete -F _ptxdist_completion p 4.) Restart your bash .. enjoy Examples: [schenk:~/projects/OSELAS.BSP-Eckelmann] eckelmann/prototyp/systemA_sd_systemB_nand* ± ./p compile [TAB][TAB] Display all 160 possibilities? (y or n) [schenk:~/projects/OSELAS.BSP-Eckelmann] eckelmann/prototyp/systemA_sd_systemB_nand* ± ./p compile ./p drop host-rauc. [TAB][TAB] host-rauc.compile host-rauc.extract host-rauc.get host-rauc.install host-rauc.prepare Regards Gavin Schenk Eckelmann AG Vorstand: Dipl.-Ing. Peter Frankenbach (Sprecher) Dipl.-Wi.-Ing. Philipp Eckelmann Dr.-Ing. Marco Münchhof Dr.-Ing. Frank Uhlemann Vorsitzender des Aufsichtsrats: Hubertus G. Krossa Stv. Vorsitzender des Aufsichtsrats: Dr.-Ing. Gerd Eckelmann Sitz der Gesellschaft: Berliner Str. 161, 65205 Wiesbaden, Amtsgericht Wiesbaden HRB 12636 http://www.eckelmann.de ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hei hei, Am Montag, 11. September 2017, 11:25:53 schrieb Guillermo Rodriguez Garcia: > The main good thing about Yocto is that it is becoming hugely popular > and has a > very strong and active community. It would be great if the same would > happen with > ptxdist :-) FWIW: somewhere in between is buildroot [1] with a community bigger than ptxdist and a lot more packages, but concepts more like ptxdist than yocto, like build stages. I liked the buildroot documentation very much, when I made my first package for it. Greets Alex [1] https://buildroot.org/ ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hei hei, Am Montag, 11. September 2017, 07:14:00 schrieb Schenk, Gavin: > * No Tab-completition as PTXDIST offers it. 5 > Stages, ~130packages and 100.000 commands. I think a very usefull > feature in PTXDIST! How does this work? I use zsh and have no tab completion for ptxdist, but miss it every now and then. :-/ Greets Alex ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hello,
My team is using ptxdist for several projects on several different boards
and while
there are of course things that can be improved, we are very happy with it.
I have not used Yocto but have looked into it from time to time; my feeling
is that
it raises the level of abstraction too much ("too much" for my personal
preferences,
that is), thus making it difficult to see what is happening under the hood.
I share
your feeling that Yocto is becoming "the Ubuntu of the embedded world".
The main good thing about Yocto is that it is becoming hugely popular and
has a
very strong and active community. It would be great if the same would
happen with
ptxdist :-)
Guillermo
2017-09-09 15:03 GMT+02:00 Erwin Rol :
> Hey all,
>
> yes I know the topic is blasphemy ;-) but I was wondering if there are
> ppl that made usecase comparisons between ptxdist and Yocto? My personal
> experience with small systems (no GUI/X/wayland) is that ptxdist is
> really easy to get going. With large systems that need QT 5.9,
> webkitgtk/chromium, different OpenGL versions etc. is that ptxdist is
> not always that easy. But I am not experienced enough with Yocto to say
> if that will be easier/better.
>
> So hence the question, anybody out there that tried both and picked
> ptxdist (or yocto) for a reason ?
>
> - Erwin
>
>
>
> ___
> ptxdist mailing list
> ptxdist@pengutronix.de
--
Guillermo Rodriguez Garcia
guille.rodrig...@gmail.com
___
ptxdist mailing list
ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hi, > > yes I know the topic is blasphemy ;-) but I was wondering if there are > ppl that made usecase comparisons between ptxdist and Yocto? My personal > experience with small systems (no GUI/X/wayland) is that ptxdist is > really easy to get going. With large systems that need QT 5.9, > webkitgtk/chromium, different OpenGL versions etc. is that ptxdist is > not always that easy. But I am not experienced enough with Yocto to say > if that will be easier/better. > > So hence the question, anybody out there that tried both and picked > ptxdist (or yocto) for a reason ? > My company uses PTXDIST for quite some Years and we go very good with it! Thanks for this great tool. In September 2016 I visited a Yocto workshop to have a look at it and build an opinion: Switch to Yocto, or stay with PTXDIST in the near future? As a result I recommended to stick to PTXDIST, and have a look at Yocto from time to time because of the big community and its progress. In September 2016 ( Maybe the information is a little outdated ) I noted: PRO Yocto: * Releasetrain every half year a new YOCTO. PTXDIST releases more often. Sometimes with only some bugfixes, sometimes major changes. Maybe a factor for planning stuff. * Fully implemented in Python you can see/change everything. (Same concept in PTXDIST). * Strict object-oriented design in the recipes. * Licensing is part of the concept in Yocto. (Meanwhile PTXDIST support license reports, too) * Documentation is extensive. (Meanwhile PTXDIST docu is getting better, too) * Package based logfiles * Reduce build time with state-caches (PTXDIST supports pre build packages, ccache and ICECREAM to reduce build time, too) * In YOCTO you have free choice for package management (RPM, DEB, ipkg, opkg). That impressed me, as I remember. And the trend in PTXDIST is to manage and update the whole firmware using RAUC, right? * The build process splits up in tasks, and for every task YOCTO creates scripts that can be called manually. Useful for searching problems I guess. (PTXDIST has its get, extract, prepare, compile, install, targetinstall stages) CON Yocto: * No nativ support for NFS boot (Maybe outdated?). The lecturer tried to sell it as an advantage, but I did not hear him ... I was in stage of shock. * Strange in such a modern systems to build semantic on top of filenames (e.g. No '_' in package name because the must only be one '_' that separates name from version ... * Changes in source (that must be marked see GPL...) are done in patchstacks (same in PTXDIST). But there is even less support for this in Yocto than PTXDIST offers (git ptx-patches). Everything has to be done manualy. For me THE big regress. * No Tab-completition as PTXDIST offers it. 5 Stages, ~130packages and 100.000 commands. I think a very usefull feature in PTXDIST! * The generic design of Yocto leaves you with a complicated, not really intuitive, folder structure. (Maybe a learning process) * No graphical support, not even dialog based. In the lesson it seemed to me that you edit configfiles until the end of time... Ok, so my conclusion was not 100% clear. Yocto is new, shiny, generic ... and not simple. Hope my information is useful Regards Gavin Schenk Eckelmann AG Vorstand: Dipl.-Ing. Peter Frankenbach (Sprecher) Dipl.-Wi.-Ing. Philipp Eckelmann Dr.-Ing. Marco Münchhof Dr.-Ing. Frank Uhlemann Vorsitzender des Aufsichtsrats: Hubertus G. Krossa Stv. Vorsitzender des Aufsichtsrats: Dr.-Ing. Gerd Eckelmann Sitz der Gesellschaft: Berliner Str. 161, 65205 Wiesbaden, Amtsgericht Wiesbaden HRB 12636 http://www.eckelmann.de ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
On Sun, 2017-09-10 at 21:24 +0200, Enrico Weigelt, metux IT consult wrote: > On 09.09.2017 18:33, Erwin Rol wrote: > > But when you want to do edit->compile->update iterations the "startup" > > time for bitbake seems to be a killer. With ptxdist I just call > > "./ptxdist compile bla" and I have no noticeable delay compared to > > native compiling. That is something I extensively use with ptxdist, and > > for Yocto that doesn't work (at least I don't know how) :-/ > > Exactly. I need incremental builds almost all the day, so yocto would > just slow me down by magnitudes. > > At that point I really wonder whether there's some really good pro > argument that might compensate that big con. I think Yocto assumes you develop your applications with the exported SDK and than just build a final distribution when the application is stable. The work flow seems very different from ptxdist. A bit like you don't rebuild Redhat to test a linux program. - Erwin ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
On 09.09.2017 18:33, Erwin Rol wrote: Exactly my problem, Yocto is becoming the Ubuntu of embedded Linux, and > like Ubuntu not everything is good about it. apropos Ubuntu: just had a ugly problem w/ my trusty notebook, which ate up my precious time: dbus didn't come up (just because a dumb file - /etc/machine-id - was missing), and therefore nothing else (except plain console) worked. oh, and some of their kernel versions suddenly miss wifi drivers, etc, etc ... But vendors like Intel, > Xilinx, TI, etc. don't even know ptxdist exists, they all only offer> some Yocto layers. chip vendors usually aren't the right parties for bsp's anyways, leave that to board vendors, or the folks here :p And the whole bitbake machinery is really slow. > > Over all I don't think bitbake is much slower than ptxdist, when> building from 0 to 100%. I'm talking about the engine itself. It always took me several minutes of metadata processing before it even starts jobs. Especially while hacking on some code, it's (IMHO) a big con if it metadata stuff sometimes takes even longer than the actual build. But when you want to do edit->compile->update iterations the "startup" time for bitbake seems to be a killer. With ptxdist I just call "./ptxdist compile bla" and I have no noticeable delay compared to native compiling. That is something I extensively use with ptxdist, and for Yocto that doesn't work (at least I don't know how) :-/ Exactly. I need incremental builds almost all the day, so yocto would just slow me down by magnitudes. At that point I really wonder whether there's some really good pro argument that might compensate that big con. --mtx ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
On Sat, Sep 09, 2017 at 08:36:13PM +0200, Tim Sander wrote: > I really would like to use ptxdist for my next private project where i > would like to try ROS. ROS1 or ROS2? > As far as i see, its already packaged for OpenEmbedded or Angstrom. I only know of old ROS packages made by the BMW folks, but this is neither current versions of ROS, nor ROS2. Do you know of newer code? > But as it has its own build tools and lots of python module > dependencies it might be to much to get all this stuff build with > ptxdist. We started with ptxdist packaging for ROS2 two techweeks ago, but it is a slow process if you have only one week per year. If someone would be interested commercially, it would probably be possible to get a solution faster, bug that didn't happen so far. > As for the speed i would say that one of my biggest gripes with ptxdist is > that -je16 -ji16 builds fail for me. And with that many virtual cores > available its realy a pitty :-(. Do you have logfiles for the fails? We are doing a lot of test builds here, and if something breaks, we would like to know. rsc -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0| Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917- | ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hi, On Sat, Sep 09, 2017 at 03:03:39PM +0200, Erwin Rol wrote: > yes I know the topic is blasphemy ;-) but I was wondering if there are > ppl that made usecase comparisons between ptxdist and Yocto? Well, it's not blasphemy at all. At Pengutronix, we are doing both, depending on the usecase and on customer requirements. The main arguments have already been brought up in this thread. > My personal experience with small systems (no GUI/X/wayland) is that > ptxdist is really easy to get going. With large systems that need QT > 5.9, webkitgtk/chromium, different OpenGL versions etc. is that > ptxdist is not always that easy. But I am not experienced enough with > Yocto to say if that will be easier/better. Like with all complicated technical tasks, things are usually easy if somebody else has already done the job and you just need to reproduce that. Yocto has many users, so quite a lot of things have already been done by somebody. If you reach the point where you need to fix things yourself, ptxdist is often easier, because Yocto/oe have quite some complexity. I suppose the rest of the story is personal preferences. When I started ptxdist, having a tool which made it easy for a developer to concentrate on his own tasks was very important to me. This is why ptxdist is focussed on solving a deverloper's tasks (not necessarily an enduser one's). But of course this is all a matter of taste, and we continue to support both. rsc -- Pengutronix e.K. | | Industrial Linux Solutions | http://www.pengutronix.de/ | Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0| Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917- | ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hi Erwin Am Samstag, 9. September 2017, 15:03:39 CEST schrieb Erwin Rol: > yes I know the topic is blasphemy ;-) but I was wondering if there are > ppl that made usecase comparisons between ptxdist and Yocto? My personal > experience with small systems (no GUI/X/wayland) is that ptxdist is > really easy to get going. With large systems that need QT 5.9, > webkitgtk/chromium, different OpenGL versions etc. is that ptxdist is > not always that easy. But I am not experienced enough with Yocto to say > if that will be easier/better. > > So hence the question, anybody out there that tried both and picked > ptxdist (or yocto) for a reason ? I really would like to use ptxdist for my next private project where i would like to try ROS. As far as i see, its already packaged for OpenEmbedded or Angstrom. But as it has its own build tools and lots of python module dependencies it might be to much to get all this stuff build with ptxdist. So i might go the OpenEmbedded road even if i have already a nice ptxdist project for the Atlas board i am using. Btw. if some is interested in this Atlas board ptxdist can publish it somewhere. As for the speed i would say that one of my biggest gripes with ptxdist is that -je16 -ji16 builds fail for me. And with that many virtual cores available its realy a pitty :-(. Best regards Tim ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
Hey all, On Sat, 2017-09-09 at 17:54 +0200, Enrico Weigelt, metux IT consult wrote: > On 09.09.2017 15:03, Erwin Rol wrote: > > but I was wondering if there are > ppl that made usecase comparisons > > between ptxdist and Yocto? My personal > > I'm currently in process of bringing up a board, where the vendor > only supplied an image built via yocto (based on an phycore-imx6, > so they just phycore's repos). Exactly my problem, Yocto is becoming the Ubuntu of embedded Linux, and like Ubuntu not everything is good about it. But vendors like Intel, Xilinx, TI, etc. don't even know ptxdist exists, they all only offer some Yocto layers. > Finding out what kernel and bbx trees it uses, and the exact configs, > and how to images are built, turned out to be quite time consuming. Yeah the learning curve is steep, but that should not be a show stopper. > And the whole bitbake machinery is really slow. Over all I don't think bitbake is much slower than ptxdist, when building from 0 to 100%. But when you want to do edit->compile->update iterations the "startup" time for bitbake seems to be a killer. With ptxdist I just call "./ptxdist compile bla" and I have no noticeable delay compared to native compiling. That is something I extensively use with ptxdist, and for Yocto that doesn't work (at least I don't know how) :-/ - Erwin ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] Yocto
On 09.09.2017 15:03, Erwin Rol wrote: Hi, yes I know the topic is blasphemy ;-) burn the witch ! ;-) but I was wondering if there are > ppl that made usecase comparisons between ptxdist and Yocto? My personal I'm currently in process of bringing up a board, where the vendor only supplied an image built via yocto (based on an phycore-imx6, so they just phycore's repos). Finding out what kernel and bbx trees it uses, and the exact configs, and how to images are built, turned out to be quite time consuming. And the whole bitbake machinery is really slow. --mtx ___ ptxdist mailing list ptxdist@pengutronix.de
[ptxdist] Yocto
Hey all, yes I know the topic is blasphemy ;-) but I was wondering if there are ppl that made usecase comparisons between ptxdist and Yocto? My personal experience with small systems (no GUI/X/wayland) is that ptxdist is really easy to get going. With large systems that need QT 5.9, webkitgtk/chromium, different OpenGL versions etc. is that ptxdist is not always that easy. But I am not experienced enough with Yocto to say if that will be easier/better. So hence the question, anybody out there that tried both and picked ptxdist (or yocto) for a reason ? - Erwin ___ ptxdist mailing list ptxdist@pengutronix.de
