Re: [gentoo-dev] Autobuilds go to /experimental and to /releases only when someone actually tests them
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Sat, 27 Jul 2013 00:13:37 -0400 "Rick \"Zero_Chaos\" Farina" wrote: > -BEGIN PGP SIGNED MESSAGE- > Hash: SHA1 > > On 07/27/2013 12:08 AM, Matt Turner wrote: > > (I sent this to gentoo-releng@. Resending to gentoo-dev@ for a > > wider audience) > > > > Can we make autobuilds go to /experimental and then only move them > > to /releases when someone actually tests them? > > > > Looking at bugzilla and listening in #gentoo-releng, it's kind of > > embarrassing how often someone downloads a Live CD only to find out > > that networking is totally broken by a udev upgrade, or something to > > that effect. > > > > We don't commit version bumps straight to stable; I don't see why we > > do with release media. > > > > It's been an odd week for me agreeing with people but yeah, I > completely agree. I think we *need* to keep the autobuilds going as > often as possible to detect obvious breakage, but there is no reason > they shouldn't be marked experimental. > > The real question is, how realistic can we make a process of testing > and moving to stable? openSUSE is using [openQA] for automated testing of installation media which is pretty need something that starts a machine in KVM and then simulates the user via a keyboard events. Last year there was also a talk about it on osc12 [2] [openQA] http://openqa.opensuse.org/ [2] http://www.youtube.com/watch?v=57a9zmpA844 > - -Zero > -BEGIN PGP SIGNATURE- > Version: GnuPG v2.0.19 (GNU/Linux) > Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ > > iQIcBAEBAgAGBQJR80jxAAoJEKXdFCfdEflKhGYP/At7Xtd4bWcY1wrxl1oPYdNr > vVfgqhmnveNqwhdERcp8InGLGoCDt+O3hvSzq4kX8qJXqizxPonP9ef1hJsnz0iw > NgjMHiGiYp2NgmU6DB7X/VLH3RNF96WJMK/R2Qtk1tuN+Ftu1D6T5hP4MmTOuvta > T2CvfYGFVAPZiY9+GLAmbe1LhjwlbJ8DnhbaamA7bK1D0ZhApWtRVtjk6unu+D5w > XRG8tIDml5gUkZRVl4d9Bg1wxuMoPtOuY2ANr+RCJPRVMkexB1XCdAVzPF73EFx+ > 0Ns5TKi+vWyhzY6PElvA0xClj2wAK/enAkAmPZ8OvagnCLfmoqZUNyr4+Eupxclt > 54pFMzpdR2KntmmFqS5ZBF8Q6nxz8GDhSm0H8+d1xTKxNcwKSlaAI7JkzBByWhKt > MjFYNTVz7MD/MFpvpRt2tKg3BI6m/ZcgCQwnAJ9QjdtyhLA8/Km5+AA2tnN457V7 > qlpf+ipjDzb3G5Po1JXSMUidy8Uu6SvqHu8TwiJUy/mlKxjmPmrPPGfRpR32pWNT > d/jE6IQAmiVjXWTDDBi0uZY8oUl5H0uroLFuA+//NtmGD8DWmV1fK0PYKLjUsE4X > nWaCKn2qlF7d16mnJh1RweBjQjMmvRYutg62A3Jb9Ek9jXBIC4bYJa2VS2xoPpWy > qZv8oon/9z336E5Uvamj > =KaUO > -END PGP SIGNATURE- > -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.19 (GNU/Linux) iQEcBAEBAgAGBQJR86ybAAoJEIN+7RD5ejah990H/jHk6bDSJeRmkDwj0FLiIc3I FbzwwaTug3DWpa1xegPTOK4Mxa2Nb7aVYfs4bZsQvQYrw+3WzvHmeYhfRiBNlU/B yiVGJ5TcPQpYltrTKx/nOuEvkH9NTcv/woiuTQ/kRedZKmNAmD/iQPwbITzOgE5U vCKyHtXGv9bwQbQBlHGNDHkinasMEGNio5uK/1XMjSxeRS9xmcAn1UNDc4OucEpA ac9EwrJgIzWgVnqD9x/mGY+GwB+dFZWrVyeGlsfEyHrTkDGu7zMBwJPd7swCxGFT ABLqts0EjbPnZIbihe962Tt/E73srEvgoHACib4D7P4sjB+X4leCQMu3D3RQVNs= =yUYu -END PGP SIGNATURE-
Re: [gentoo-dev] Autobuilds go to /experimental and to /releases only when someone actually tests them
On Sat, Jul 27, 2013 at 6:05 AM, Fabio Erculiani wrote: > Some time ago I was also thinking about writing a test framework for > testing live images through kvm. > Of course I didn't manage to find time to try to arrange something in > the end, but the idea is still popping up in my mind every now and > then. Feel free to adapt: https://github.com/rich0/rich0-gentoo-bootstrap FYI - I plan to update it slightly today - I'm actually running 4 builds right now, eliminating some config-file cruft that is left over from earlier failures, and adding a little more due to a new one. More often than not I run into a little surprise every time I run it, either due to regressions, or due to the stage3 being so old that half the system needs to be updated. Newer stage3s would certainly help. So, if we add a QA step to stage3 publication we should try to keep up the pace, otherwise we'll create just as many problems for new users having to use old stage3s as we'll solve. Those scripts could benefit from re-factoring (oh, and it wouldn't hurt for me to credit Dowd and Associates again who created them). Some thoughts: 1. I added the plugin framework but never moved some of the default packages into a plugin. That would be an easy improvement to improve utility. 2. All of the EC2 logic itself could be moved into appropriate functions, and then the script could be made configurable for EC2 vs KVM vs whatever. The bootstrap logic itself would work fine in an architecture once /mnt/gentoo is setup. I wouldn't necessarily mind volunteering to keep up with it, but if I'm going to be running these weekly I'll probably ask for reimbursement for EC2 expenses (I use spot instances so it isn't too much, but all that CPU still adds up as well as storage costs while I'm debugging things). If we're willing to pay for the storage (fairly nominal) I could also work on publishing a list of up-to-date EC2 images (they're a byproduct of this stuff anyway so I just need to publish them when they work). Rich
Re: [gentoo-dev] Autobuilds go to /experimental and to /releases only when someone actually tests them
Some time ago I was also thinking about writing a test framework for testing live images through kvm. Of course I didn't manage to find time to try to arrange something in the end, but the idea is still popping up in my mind every now and then. -- Fabio Erculiani
Re: [gentoo-dev] Autobuilds go to /experimental and to /releases only when someone actually tests them
Le vendredi 26 juillet 2013 à 21:19 -0700, "Paweł Hajdan, Jr." a écrit : > > The real question is, how realistic can we make a process of testing and > > moving to stable? > > We have arch teams, we have users... when several users say it's OK I > think it is OK. As compared to a script pushing it to the website just > because it compiled. How about a regular "test livecds" event? Kind of like a bug day. Seems like it could also be a nice and easy way to reach potential new contributers: just burn an ISO, reboot your computer, report back whether network/disk access works. Rémi
Re: [gentoo-dev] Autobuilds go to /experimental and to /releases only when someone actually tests them
On 7/26/13 9:13 PM, Rick "Zero_Chaos" Farina wrote: > On 07/27/2013 12:08 AM, Matt Turner wrote: >> Can we make autobuilds go to /experimental and then only move them to >> /releases when someone actually tests them? Very interesting. :) I had a similar idea. I think it's great. >> Looking at bugzilla and listening in #gentoo-releng, it's kind of >> embarrassing how often someone downloads a Live CD only to find out >> that networking is totally broken by a udev upgrade, or something to >> that effect. Yes - and it's very important to make that first experience with the distro as good as possible. The bugs are usually not fixed quickly enough anyway. I'd like to add a suggestion - document the processes better and allow more people to contribute. >> We don't commit version bumps straight to stable; I don't see why we >> do with release media. > > It's been an odd week for me agreeing with people but yeah, I completely > agree. I think we *need* to keep the autobuilds going as often as > possible to detect obvious breakage, but there is no reason they > shouldn't be marked experimental. +1 > The real question is, how realistic can we make a process of testing and > moving to stable? We have arch teams, we have users... when several users say it's OK I think it is OK. As compared to a script pushing it to the website just because it compiled. Paweł signature.asc Description: OpenPGP digital signature
Re: [gentoo-dev] Autobuilds go to /experimental and to /releases only when someone actually tests them
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 07/27/2013 12:08 AM, Matt Turner wrote: > (I sent this to gentoo-releng@. Resending to gentoo-dev@ for a wider audience) > > Can we make autobuilds go to /experimental and then only move them to > /releases when someone actually tests them? > > Looking at bugzilla and listening in #gentoo-releng, it's kind of > embarrassing how often someone downloads a Live CD only to find out > that networking is totally broken by a udev upgrade, or something to > that effect. > > We don't commit version bumps straight to stable; I don't see why we > do with release media. > It's been an odd week for me agreeing with people but yeah, I completely agree. I think we *need* to keep the autobuilds going as often as possible to detect obvious breakage, but there is no reason they shouldn't be marked experimental. The real question is, how realistic can we make a process of testing and moving to stable? - -Zero -BEGIN PGP SIGNATURE- Version: GnuPG v2.0.19 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQIcBAEBAgAGBQJR80jxAAoJEKXdFCfdEflKhGYP/At7Xtd4bWcY1wrxl1oPYdNr vVfgqhmnveNqwhdERcp8InGLGoCDt+O3hvSzq4kX8qJXqizxPonP9ef1hJsnz0iw NgjMHiGiYp2NgmU6DB7X/VLH3RNF96WJMK/R2Qtk1tuN+Ftu1D6T5hP4MmTOuvta T2CvfYGFVAPZiY9+GLAmbe1LhjwlbJ8DnhbaamA7bK1D0ZhApWtRVtjk6unu+D5w XRG8tIDml5gUkZRVl4d9Bg1wxuMoPtOuY2ANr+RCJPRVMkexB1XCdAVzPF73EFx+ 0Ns5TKi+vWyhzY6PElvA0xClj2wAK/enAkAmPZ8OvagnCLfmoqZUNyr4+Eupxclt 54pFMzpdR2KntmmFqS5ZBF8Q6nxz8GDhSm0H8+d1xTKxNcwKSlaAI7JkzBByWhKt MjFYNTVz7MD/MFpvpRt2tKg3BI6m/ZcgCQwnAJ9QjdtyhLA8/Km5+AA2tnN457V7 qlpf+ipjDzb3G5Po1JXSMUidy8Uu6SvqHu8TwiJUy/mlKxjmPmrPPGfRpR32pWNT d/jE6IQAmiVjXWTDDBi0uZY8oUl5H0uroLFuA+//NtmGD8DWmV1fK0PYKLjUsE4X nWaCKn2qlF7d16mnJh1RweBjQjMmvRYutg62A3Jb9Ek9jXBIC4bYJa2VS2xoPpWy qZv8oon/9z336E5Uvamj =KaUO -END PGP SIGNATURE-
[gentoo-dev] Autobuilds go to /experimental and to /releases only when someone actually tests them
(I sent this to gentoo-releng@. Resending to gentoo-dev@ for a wider audience) Can we make autobuilds go to /experimental and then only move them to /releases when someone actually tests them? Looking at bugzilla and listening in #gentoo-releng, it's kind of embarrassing how often someone downloads a Live CD only to find out that networking is totally broken by a udev upgrade, or something to that effect. We don't commit version bumps straight to stable; I don't see why we do with release media. Matt