Re: [gentoo-user] Portage.provided
On Jun 20, 2015, at 12:43, Franz Fellner alpine.art...@gmail.com wrote: Matti Nykyri wrote: How to get portage off my back? I have the following in /etc/portage/package.provided: For me package.provided didn't work wither. Until I noticed that I missed profile in te path. mv /etc/portage/package.provided /etc/portage/profile/package.provided and it should do what you expect Yeap. Sorry. Too fast reading of man pages :) Now it works after I moved the file. sys-kernel/gentoo-sources-4.0.5 sys-kernel/gentoo-sources-4 sys-kernel/gentoo-sources However when I run emerge -DuvaN world: These are the packages that would be merged, in order: Calculating dependencies... done! [ebuild N ] sys-kernel/gentoo-sources-4.0.5:4.0.5::gentoo USE=-build -deblob -experimental -symlink 0 KiB Total: 1 package (1 new), Size of downloads: 0 KiB Would you like to merge these packages? [Yes/No] What should I do? -- -Matti
Re: [gentoo-user] Blank screen after hibernation with radeon driver
On Jun 20, 2015, at 5:16, Fernando Rodriguez frodriguez.develo...@outlook.com wrote: Hello, After switching from fglrx to the radeon driver I get a blank screen after resuming from hibernation. I can ssh in but I can't restart xorg. This happens with pm-utils and also with systemd. Suspend works fine with both. My video card is: VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Kabini [Radeon HD 8210] Any suggestions? Ssh in and run commands: export DISPLAY=:0 xrandr Or xrandr --output --mode etc with right arguments If that doesn't help, use top and perf top to see what is going on. Does restarting X fix te issue? -- -Matti
[gentoo-user] Portage.provided
How to get portage off my back? I have the following in /etc/portage/package.provided: sys-kernel/gentoo-sources-4.0.5 sys-kernel/gentoo-sources-4 sys-kernel/gentoo-sources However when I run emerge -DuvaN world: These are the packages that would be merged, in order: Calculating dependencies... done! [ebuild N ] sys-kernel/gentoo-sources-4.0.5:4.0.5::gentoo USE=-build -deblob -experimental -symlink 0 KiB Total: 1 package (1 new), Size of downloads: 0 KiB Would you like to merge these packages? [Yes/No] What should I do? -- -Matti
RE: [gentoo-user] Portage.provided
Matti Nykyri wrote: How to get portage off my back? I have the following in /etc/portage/package.provided: For me package.provided didn't work wither. Until I noticed that I missed profile in te path. mv /etc/portage/package.provided /etc/portage/profile/package.provided and it should do what you expect. sys-kernel/gentoo-sources-4.0.5 sys-kernel/gentoo-sources-4 sys-kernel/gentoo-sources However when I run emerge -DuvaN world: These are the packages that would be merged, in order: Calculating dependencies... done! [ebuild N ] sys-kernel/gentoo-sources-4.0.5:4.0.5::gentoo USE=-build -deblob -experimental -symlink 0 KiB Total: 1 package (1 new), Size of downloads: 0 KiB Would you like to merge these packages? [Yes/No] What should I do? -- -Matti
[gentoo-user] Jenkins
Hello one and all, I want to first install Jenkins on a single multicore amd system, so I found this brief guide (which seems simple enough): https://code.google.com/p/godin-gentoo-repository/wiki/Jenkins What I also need is an example collection of code to run Jenkins-CI scripts on, sort of as a benchmark. I figure it would be best to learn about Jenkins by merely trying to follow/duplicate an existing (hopefully small) Jenkins CI project. Basically a monkey-see-monkey-duplicate sort of endeavour. This will also allow me to then duplicate the Jenkins-CI onto the gentoo mesos clusters (small right now) I have built to debug problems or issues with Jenkins running on a gentoo-mesos-cluster. Ideas on how to perform comparison (benchmarks) between the single multi-processor Jenkins-CI and the one on top of a cluster, would be warmly received too. After that is completed and stable, then I'm going to attempt to move other codes to this gentoo-mesos-jenkins-CI framework. Any suggestions are most welcome. James
Re: [gentoo-user] Jenkins
On Sat, Jun 20, 2015 at 07:03:18PM +, James wrote: Hello one and all, I want to first install Jenkins on a single multicore amd system, so I found this brief guide (which seems simple enough): https://code.google.com/p/godin-gentoo-repository/wiki/Jenkins I highly recommend playing with the Jenkins docker container to get a feel for using the web UI: docker run -p 8080:8080 -d jenkins The official jenkins image comes with the bare minimum of libraries so good luck building any C/C++ project, but it is helpful just to point and click around (if you already have docker, that is). What I also need is an example collection of code to run Jenkins-CI scripts on, sort of as a benchmark. I figure it would be best to learn about Jenkins by merely trying to follow/duplicate an existing (hopefully small) Jenkins CI project. Basically a monkey-see-monkey-duplicate sort of endeavour. This will also allow me to then duplicate the Jenkins-CI onto the gentoo mesos clusters (small right now) I have built to debug problems or issues with Jenkins running on a gentoo-mesos-cluster. Ideas on how to perform comparison (benchmarks) between the single multi-processor Jenkins-CI and the one on top of a cluster, would be warmly received too. This is not really a fair comparison. Jenkins has no concept of cores, only executors. Each executor can run one build at a time. Jenkins was written with Java in mind (AFAIK), and Java build tools are the reason for this. `javac` by default launches some number of threads (not sure how many on other systems, but on my laptop it's usually 3) and compiles in parallel. I'm not sure it's possible to control this (someone feel free to correct me if I'm wrong), hence Jenkins does not care about cores and only about how many builds are allowed to run in parallel at once. This is different than make and just other build tools in general that I've worked with for other languages. So, to the point; when you add more build slaves with more executors, you will be able to run more builds in parallel and it will therefore be faster. This gets even more complicated since you can constrain builds to only run on certain hosts and various other complicated setups, but in general more executors means more builds in parallel means faster. After that is completed and stable, then I'm going to attempt to move other codes to this gentoo-mesos-jenkins-CI framework. Any suggestions are most welcome. James My main suggestion is to not bother running a Jenkins cluster and stick with a single host setup unless: * You are building so many things that utilization is near 100% * You are targeting multiple platforms and therefore need multiple build hosts Alec P.S. I could not find any reference to parallel, threads, processes or anything else on javac's man page that makes me think it is easy to set the number of threads it uses.
[gentoo-user] Re: Jenkins
Alec Ten Harmsel alec at alectenharmsel.com writes: On Sat, Jun 20, 2015 at 07:03:18PM +, James wrote: Hello one and all, I want to first install Jenkins on a single multicore amd system, so I found this brief guide (which seems simple enough): https://code.google.com/p/godin-gentoo-repository/wiki/Jenkins I highly recommend playing with the Jenkins docker container to get a feel for using the web UI: docker run -p 8080:8080 -d jenkins The official jenkins image comes with the bare minimum of libraries so good luck building any C/C++ project, but it is helpful just to point and click around (if you already have docker, that is). Yes, I've used Docker on gentoo before. NO, it's currently un-installed and using docker for what I am after only complicates things at this point. This is not really a fair comparison. Jenkins has no concept of cores, only executors. Each executor can run one build at a time. Jenkins was written with Java in mind (AFAIK), and Java build tools are the reason for this. `javac` by default launches some number of threads (not sure how many on other systems, but on my laptop it's usually 3) and compiles in parallel. I'm not sure it's possible to control this (someone feel free to correct me if I'm wrong), hence Jenkins does not care about cores and only about how many builds are allowed to run in parallel at once. I'll have to drill into javac a bit more, it seems. There is jenkins-bin on gentoo; it should (?) be a quick install and config. I'll keep in mind what you are saying, but my main goal is to find (CI) codes that I can run on gentoo-mesos clusters. Nothing about Jenkins is a critical need for me. Although, as a consultant I do routinely get asked about Jenkins experience, so just noodling around with it a bit is a good idea for me. So, to the point; when you add more build slaves with more executors, you will be able to run more builds in parallel and it will therefore be faster. This gets even more complicated since you can constrain builds to only run on certain hosts and various other complicated setups, but in general more executors means more builds in parallel means faster. AT some point, on a single multiprocessor system, Jenkins will slow down even it more resources are configured for it (overall system load, if nothing else)? AT that point, I can note the resources and apply those limits to the cluster and note the performance differences. Surely, a cluster with more resources will be able to do more (CI_Jenkins) work; so discovering those details is a primary goal of this endeavour. My main suggestion is to not bother running a Jenkins cluster and stick with a single host setup unless: Sorry, the cluster, is the main_goal, not Jenkins or CI. Buds at cisco are all very braggadocios about CI (Jenkins and others) on their mesos clusters. * You are building so many things that utilization is near 100% * You are targeting multiple platforms and therefore need multiple build hosts YES, that is exactly the ultimate goal. A CI cluster, open to friends, pals and customers (maybe). It's all about the mesos cluster, you should know that. P.S. I could not find any reference to parallel, threads, processes or anything else on javac's man page that makes me think it is easy to set the number of threads it uses. No surprise there. Configuring threads, cgroups, and memory resources is a hotly contested area of the cluster codes I have (am) working with. Forget bikeshedding, it Blood_shedding... especially now that RDMA is hotly being pursued by the many. Thanks for your insights, James
[gentoo-user] Re: Profile listings
-BEGIN PGP SIGNED MESSAGE- Hash: SHA512 On 2015-06-19 15:46, James wrote: Martin Vaeth martin at mvath.de writes: James wireless at tampabay.rr.com wrote: # PORTAGE_PROFILE=/usr/portage/profiles/arch/arm/armv7a eix -c --system No matches found. Obviously, this profile contains no at system packages. Which appears natural for an embedded profile... Obviously, one cannot obtain the profiles to other arches from the data found in /usr/portage/profile, easily. Surely a front-end would be keen for this. Also, I had a friend on an embedded gentoo (arm) board verify that the same 42 files for @system was installed on his arm board (eix -e --system). I surely hope that something (gui tool) convenient and robust becomes available; maybe GLEP64 will help. For embedded (any arch) I would expect that the @system would not contain all the files necessary to compile code. After all, that's really what cross-compiling is all about. I'm not sure a single packages, such as busybox really contains the best/complete codes that is needed on an embedded gentoo system, but that is a different issue. I also think there is room for another profile, between default and embedded where the target is a single (or focused) build for something like a sniffer, a data collector, a firewall, a bridge, a router, etc etc to have less than the default profile and specifically matched to a tuned (aggressively pruned) kernel for a very specific and limited purpose. That said, I'm going to think about this a bit more and marinate over the postings from Andreas and others for a while longer to decide what I think it should really be. I also think there should be a well defined path of what and how to migrate from embedded to minimized[focused] and default systems. One could experiment for example experiment with running a gentoo based firewall-router on an embedded gentoo system, a minimized[focused] gentoo system and a default profile gentoo system all with the same firewall-routers codes for cost and security and performance evaluations. Thanks to all for the excellent information and input! Sorry about being dense, as now Andreas's posts make more sense, but also highlight the shortness of breadth of gentoo's current profile system. It's also a pig mess of code, ideas and old constructs, imho. (note: nothing negative about the wonderful folks that have maintained and extended profiles over the years, but, it is time for a discussion and new architecture for the entire profile landscape, imho. Maybe after Glep 64 is usable it would be a good time to move forward on profile_modernizations.. Others comments are welcome. James The list of all profiles that can be chosen (for all architectures) can be found in ${PORTDIR}/profiles/profiles.desc . There are other profile-like directories under ${PORTDIR}/profiles, but these are only used as parents for a complete profile, as would be listed in profiles.desc. Most profiles do not change much, if anything, in the @system set. The @system set contains much more than you would probably need for a dedicated, embedded device. - -- Jonathan Callen -BEGIN PGP SIGNATURE- Version: GnuPG v2 iQIcBAEBCgAGBQJVhi96AAoJEEIQbvYRB3mggLMP/3xi0EQFcHXx1rrPufYq/is4 VVne2H9PtvFtfhqCVpjqIMFqknL0XwtjDRJx/EdFO1Ym02tR5OX/iU4hcmyRVg9X 6OgecksDhtQVs4UVfNkjBOEbMUMMFKEimboLq1w9j8RwmjMx84ZYuhkNag33d72X 4St4ly8Y7w1feeirn925of7Dj7upQeievpDs6kK7WtLIA8t8nZeBmNFUfkjlAfCe YSukUBqzK8vq92M5jmJRbtPaOePZppJBRcmiPOqqF8uhtXozo9dgoOk1TANXtNEV fip4NLczVM8eOf54JLAM6ttBuBK1yTQx4csnPBbd6WU3vD2E0YSuZjFADFBWsTiY 8Q7ZvZIg7i60ZwzQ127MTgOQQDYHEgpEorWC9X1EKHMIke6k9mFQtdaGMPIkb8jt 3F/LbV4YP6h0Q6QQdQq4LpWBmvZ78LmJwm5KtXMZean4Z5G3rSzmbu/nsSJy0zEW zJu2vKcitzzJNE7c0CBpWVUcUj9ZB819ao5tMxbft/LJNTgURz7wScW1FSS6R+n1 EzQgBQdWyIXaYMqAAapYrMgZhKdij4NAGp7rUi+uIIrxleu5ECh9a6/VfVr9Z7V8 v+uLYuiBX5agFvjA7UCy5gq/6vD/QmlWlh88lMpp0dBLTN/ovM3CcBH5h0rOBHxf Z7gzy0i/uEhZoo235pJc =9peY -END PGP SIGNATURE-
Re: [gentoo-user] Blank screen after hibernation with radeon driver
On Saturday, June 20, 2015 10:15:37 AM Matti Nykyri wrote: On Jun 20, 2015, at 5:16, Fernando Rodriguez frodriguez.develo...@outlook.com wrote: Hello, After switching from fglrx to the radeon driver I get a blank screen after resuming from hibernation. I can ssh in but I can't restart xorg. This happens with pm-utils and also with systemd. Suspend works fine with both. My video card is: VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Kabini [Radeon HD 8210] Any suggestions? Ssh in and run commands: export DISPLAY=:0 xrandr Or xrandr --output --mode etc with right arguments If that doesn't help, use top and perf top to see what is going on. Does restarting X fix te issue? I can't restart X, it hangs when I try, xrandr also hangs. Even kill -9 doesn't kill it. The logs sometimes don't show anything at all but today it's show this consistently: un 20 18:19:08 navi kernel: [drm:cik_ring_test] *ERROR* radeon: ring 1 test failed (scratch(0x3010C)=0xCAFEDEAD) Jun 20 18:19:08 navi kernel: ata1: SATA link up 6.0 Gbps (SStatus 133 SControl 300) Jun 20 18:19:08 navi kernel: ata1.00: configured for UDMA/133 Jun 20 18:19:08 navi kernel: [drm:cik_ring_test] *ERROR* radeon: ring 2 test failed (scratch(0x3010C)=0xCAFEDEAD) Jun 20 18:19:08 navi kernel: [drm] ring test on 3 succeeded in 4 usecs Jun 20 18:19:08 navi kernel: [drm] ring test on 4 succeeded in 4 usecs Jun 20 18:19:08 navi kernel: [drm] ring test on 5 succeeded in 1 usecs Jun 20 18:19:08 navi kernel: [drm] UVD initialized successfully. Jun 20 18:19:08 navi kernel: [drm] ring test on 6 succeeded in 812 usecs Jun 20 18:19:08 navi kernel: [drm] ring test on 7 succeeded in 3 usecs Jun 20 18:19:08 navi kernel: [drm] VCE initialized successfully. Jun 20 18:19:08 navi kernel: [drm] ib test on ring 0 succeeded in 0 usecs Jun 20 18:19:08 navi kernel: [drm] ib test on ring 3 succeeded in 0 usecs Jun 20 18:19:08 navi kernel: [drm] ib test on ring 4 succeeded in 0 usecs Jun 20 18:19:08 navi kernel: [drm] ib test on ring 5 succeeded Jun 20 18:19:08 navi kernel: [drm] ib test on ring 6 succeeded Jun 20 18:19:08 navi kernel: [drm] ib test on ring 7 succeeded Jun 20 18:19:08 navi kernel: [drm:radeon_dp_link_train_cr] *ERROR* displayport link status failed Jun 20 18:19:08 navi kernel: [drm:radeon_dp_link_train_cr] *ERROR* clock recovery failed Jun 20 18:19:08 navi kernel: [drm:radeon_dp_link_train_cr] *ERROR* displayport link status failed Jun 20 18:19:08 navi kernel: [drm:radeon_dp_link_train_cr] *ERROR* clock recovery failed On one occassion there where about 20 backtraces on the log. That log is gone but it had a message basically saying (I don't remember the exact words) that something went wrong and the kernel recovered but a reboot was still needed. I'll post if it comes up again. Also relevant that I've been getting this and similar errors on my logs for a few days (probably since I switched to the radeon driver), all the i2c devices are on the video card: Jun 20 16:46:30 navi kernel: i2c i2c-7: sendbytes: error -110 Jun 20 16:46:41 navi kernel: i2c i2c-7: sendbytes: error -110 Jun 20 16:47:01 navi kernel: i2c i2c-7: sendbytes: error -110 Jun 20 16:47:11 navi kernel: i2c i2c-7: sendbytes: error -110 Jun 20 16:47:31 navi kernel: i2c i2c-7: sendbytes: error -110 -- Fernando Rodriguez