Bug#349354: why not yaird by default?
On Mon, Feb 27, 2006 at 08:46:49PM +, Stephen Gran wrote: This one time, at band camp, friendly Sven Luther said: On Mon, Feb 27, 2006 at 03:43:17PM +0100, Jonas Smedegaard wrote: On Mon, 27 Feb 2006 15:14:02 +0100 Sven Luther [EMAIL PROTECTED] wrote: #343427: linux-image-2.6.14-2-powerpc: Installation fails #345067: [powerpc] ide-generic is not built on powerpc, yaird tries to include it and fails Both relate to ide-generic. Difference between yaird and initramfs-tools in regards to this issue is that yaird has builtin probing while initramfs-tools rely on udev for extracting kernels own logic and/or implement workarounds. What has that to do with anything ? Since both bugs are arguably kernel bugs (some modules on some platforms can't work without also loading ide-generic, but the kernel provides no mechanism to find that out), I think it has rather a lot to do with the issue at hand. Well, sure, but the maintainership of MIA is crap, and the debian maintainer is unwilling to be reasonable and leaves unable to install kernel bugs with patches open for months. We cannot consider such a package of enough quality enough to even consider it for etch unless something changes with the maintainership. This was my point, not some random technical babling about the difference between yaird and initramfs-tools (and yes, i was a fervent supporter of yaird, and strongly advocated making it the default previously, so i am aware of the technical issues). The question was should yaird not be made the default and i answered that this is probably not a good idea because the DD maintainer (you) doesn't seem able to fix bugs without consulting his upstream and that said upstream is MIA. An MIA upstream is indeed a serious problem. A maintainer being unwilling to accept a bad hack to work around brokenness elsewhere is less of an issue, at least IMHO. Well, the problem was introduced in a bad hack without any kind of understanding about the issue in the first place, the proposed problem is just desactivating the hack on powerpc, where we know we don't build the ide-generic module, so i doubt anyone can prove me it is *NEEDED* in any way. In erkelenz i disucssed this with jonas, told him let's look at this and convince ourself that it is no problem, and was only told that he would not do som, because he was not able to be sure that it would not break on some random user setup, and without getting his upstream approval. I wrote upstream immediately, but we got no feedback, this was over a month ago, and yaird remains broken. And to make things clear, if loading ide-generic on powerpc would ever be *NEEDED*, then the case of not building ide-generic would not work, and it has been working just fine. So, the issue is double, first the upstream maintainer is MIA, which is not nice, but second the debian maintainer is unable or unwilling to take his maitainer job seriously and at least consider looking at the patches that are submitted by the folk who have the hardware. This jonas clearly said (and so loudly that folk in Erkelenz asked us to leave the room) that he would not look at my patch without aproval from upstream, that he didn't really understand yaird enough to be sure that nothing else would break if he did that change (which just reverted a previously applied hacky patch that broke this), and was thus not even considering looking it over with me. In these conditions, it is unacceptable to make yaird the default (or probably even ship it with etch), if we don't get a change in maintainership, either jonas becoming more responsible, or someone co-maintaining it or taking it over, preferably someone with a clue and knowledgeable in perl. Friendly, Sven Luther Ever so friendly, -- - | ,''`.Stephen Gran | | : :' :[EMAIL PROTECTED] | | `. `'Debian user, admin, and developer | |`- http://www.debian.org | - -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#349354: why not yaird by default?
This one time, at band camp, friendly Sven Luther said: [ a long unfriendly rant snipped ] Frankly, I don't care whether or not yaird is the default ramdisk generator for the kernel. An MIA upstream is a good reason to decide against it, in fact. The entire rest of your argument sounds like hurt feelings because Jonas won't take a broken patch, though, and I'm just not interested in that sort of silliness. The fact that this same bug exist(s|ed) in udev, mkinitramfs, and yaird indicates that there is a real bug in the kernel that is merely being triggered by all of these packages. Can you, as kernel maintainer, please spend your energy fixing the kernel bug that is causing all of this, instead of wasting my time bickering with Jonas? This bug rendered my laptop unbootable until Md patched around it. Very very friendly, -- - | ,''`.Stephen Gran | | : :' :[EMAIL PROTECTED] | | `. `'Debian user, admin, and developer | |`- http://www.debian.org | - signature.asc Description: Digital signature
Bug#349354: why not yaird by default?
This one time, at band camp, friendly Sven Luther said: On Tue, Feb 28, 2006 at 10:17:39AM +, Stephen Gran wrote: This one time, at band camp, friendly Sven Luther said: [ a long unfriendly rant snipped ] Frankly, I don't care whether or not yaird is the default ramdisk generator for the kernel. An MIA upstream is a good reason to decide against it, in fact. The entire rest of your argument sounds like hurt feelings because Jonas won't take a broken patch, though, and I'm just not interested in that sort of silliness. And who are you to say it is a broken patch ? Have you looked at it ? The main problem is that jonas does not even want to look at the patch, so claiming it is broken without looking at it, it kind of insulting in the first place. Please have a look at the patch, and show me how it is broken. Of course I've looked at the patch, that's how I came to the conclusion it's broken. Please don't jump to conclusions. Your patch makes the assumption that ide-generic will never be needed on any ppc hardware ever. This may be the case for you right now, but does not appear to me to be a safe assumption. This is exactly the sort of hack that leads to more problems down the road, patched around with worse hacks then, ad nauseum. It is vastly more appropriate to fix the single real bug than to patch around it in all the places that trigger it. The fact that this same bug exist(s|ed) in udev, mkinitramfs, and yaird indicates that there is a real bug in the kernel that is merely being I personally believe that the real bug is in ide-generic. Then fix it. You are a kernel maintainer, right? triggered by all of these packages. Can you, as kernel maintainer, please spend your energy fixing the kernel bug that is causing all of this, instead of wasting my time bickering with Jonas? This bug rendered my laptop unbootable until Md patched around it. Yeah, how does you like it ? And i am in a situation where my RL work does include doing support for 1000+ pegasos users out there, whose debian install gets broken by jonas and erik's half-backed patch, and erik is MIA, and jonas is not even interested in thinking about fixing it. He never even replied to the bug report until i pointed it out in erkelenz to him. So, the thing you are complaining about, is exactly the same i am complaining about here. So, can you please fix it? Since you're a kernel team member, you are in a better position than either Jonas or myself to do something about it. Or is there some problem with fixing it the right way that I'm missing? -- - | ,''`.Stephen Gran | | : :' :[EMAIL PROTECTED] | | `. `'Debian user, admin, and developer | |`- http://www.debian.org | - signature.asc Description: Digital signature
Bug#349354: why not yaird by default?
What about you all stop Cc'ing this non-relevant bug? Thank you... (FWIW my opinion is that yaird should not be the default because hardware changes will make the system unbootable if the drivers needed to mount / are not in the initramfs.) -- ciao, Marco signature.asc Description: Digital signature
Bug#349354: why not yaird by default?
On Tue, Feb 28, 2006 at 11:17:36AM +, Stephen Gran wrote: Please have a look at the patch, and show me how it is broken. Of course I've looked at the patch, that's how I came to the conclusion it's broken. Please don't jump to conclusions. Your patch makes the assumption that ide-generic will never be needed on any ppc hardware Ok. so you believe that there may be a remote chance that ide-generic will be needed on powerpc, this would mean that there would be no ide working without it right. I mean, look into your dictionary for 'needed', as i guess it is safe to say that is something that is 'needed' is missing, then it doesn't work. Since it is working right now, it is proof enough that it is not needed. If this will change in a random future, then it will be time enough to fix it for such an hypothetical situation, so basically you are breaking a currently existing case for some hypothetical future case, how logical. ever. This may be the case for you right now, but does not appear to me to be a safe assumption. This is exactly the sort of hack that leads to Ah, yes. That is also what jonas claims. Please explaqin to me a scenario where this assumption will not be broken. I mean, face it, the 'let's include ide-generic' hack was activated for piix, via-ide, and third one i don't remember. None of them have vocation to work in the main case on powerpc hardware. I guess you could have via-ide pci cards in a powermac, but this is far from being common, compared to the 1000+ users we have out there and i have to do support for. more problems down the road, patched around with worse hacks then, ad nauseum. It is vastly more appropriate to fix the single real bug than to patch around it in all the places that trigger it. Nope, the patch just disables the ugly hack erik and jonas enabled in the first place. Notice also that even on x86 it is not clear that this hack is needed in the majority of cases, as we only had a handful of reports about this. The fact that this same bug exist(s|ed) in udev, mkinitramfs, and yaird indicates that there is a real bug in the kernel that is merely being I personally believe that the real bug is in ide-generic. Then fix it. You are a kernel maintainer, right? The whole ide layer is going to go away in favour of the new libata reimplementation, upstream is working on this, it is not quite mature enough yet. triggered by all of these packages. Can you, as kernel maintainer, please spend your energy fixing the kernel bug that is causing all of this, instead of wasting my time bickering with Jonas? This bug rendered my laptop unbootable until Md patched around it. Yeah, how does you like it ? And i am in a situation where my RL work does include doing support for 1000+ pegasos users out there, whose debian install gets broken by jonas and erik's half-backed patch, and erik is MIA, and jonas is not even interested in thinking about fixing it. He never even replied to the bug report until i pointed it out in erkelenz to him. So, the thing you are complaining about, is exactly the same i am complaining about here. So, can you please fix it? Since you're a kernel team member, you are in a better position than either Jonas or myself to do something about it. Or is there some problem with fixing it the right way that I'm missing? It is loadful of work, and it is not even clear what the problem is exactly, and we don't have access to the hardware who exhibits the problem, and what else more ... So, instead of investigating this, both the yaird and initramfs-tools have gone into doing ugly hacks, which broke the previously perfectly working pegasos system, and jonas is plainly refusing to even think about it. I even came quite friendly to him in erkelenz and said let's fix it together this WE, and what, no he prefered to go into hours of polimicking about philospohical reasons why it 'may' break in some undetermined future, and got angry at any attempt on my part to show him the code or to explain to him that there is no chance it may break. I mean he clearly decided that anything i may say is just to be ignored, so what do you want ? And since the upstream is MIA, i only see two solutions, takeover yaird, or let it be and go with initramfs-tools as default, given that i don't speak perl, and even if i think yaird is the better concept, well, there is not much choice of what to do. Friendly, Sven Luther -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#349354: why not yaird by default?
On Mon, Feb 27, 2006 at 03:01:39PM +0100, Jonas Smedegaard wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Mon, 27 Feb 2006 14:36:39 +0100 Sven Luther [EMAIL PROTECTED] wrote: The quality of the maintainers of both tool also play an important role in which tool to chose as standard, and we have a pretty good feedback from both the initramfs-tools maintainer and upstream, while for yaird, the upstream maintainer seems to be MIA since around late december, and the debian maintainer seems to be unable to fix a bug or even looking at patches without getting feedback from the upstream maintainer, which makes yaird unsuitable to be the default. ...says a single member of the kernel team. I would appreciate comments from other members of the kernel team on this issue. Please tell us since when you have not heard about your upstream, and what fixes you have down to yaird since then, and when do you expect to be looking at merged bug : #343427: linux-image-2.6.14-2-powerpc: Installation fails #345067: [powerpc] ide-generic is not built on powerpc, yaird tries to include it and fails Open respectively since 74 and 60 days now. I even proposed you to go over the patch and solve it in erkelenz, but you where more interested in polimicking than fixing this bug, what else can i say ... Sven Luther -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#349354: why not yaird by default?
On Mon, 27 Feb 2006 15:14:02 +0100 Sven Luther [EMAIL PROTECTED] wrote: #343427: linux-image-2.6.14-2-powerpc: Installation fails #345067: [powerpc] ide-generic is not built on powerpc, yaird tries to include it and fails Both relate to ide-generic. Difference between yaird and initramfs-tools in regards to this issue is that yaird has builtin probing while initramfs-tools rely on udev for extracting kernels own logic and/or implement workarounds. So the maintainer of initramfs-tools is right in not having any hanging bugreports about ide-generic (they should be pushed either to udev or linux-2.6) while yaird do not have such luxury. If I remember correctly then the maintainer of udev has been quite sceptical about implementing workarounds for the ide-generic problem (rather than considering it a kernel bug) too. I am still interested in opinions from _other_ members of the kernel team: Is it decided to decided to ditch yaird due to my irresponsible or non-cooperative behaviour (which I claim is related to the problematic ide-generic issue), or do you have different opinions on how ramdisk generators should be picked for the official kernels? Yes, I am perfectly aware that not all is in favor of yaird, but the reasoning behind choosing one over the other - or what is considered fair ground for fair a competition between them. - Jonas -- * Jonas Smedegaard - idealist og Internet-arkitekt * Tlf.: +45 40843136 Website: http://dr.jones.dk/ - Enden er nær: http://www.shibumi.org/eoti.htm pgpb1cBSYvjkO.pgp Description: PGP signature
Bug#349354: why not yaird by default?
On Mon, Feb 27, 2006 at 03:43:17PM +0100, Jonas Smedegaard wrote: On Mon, 27 Feb 2006 15:14:02 +0100 Sven Luther [EMAIL PROTECTED] wrote: #343427: linux-image-2.6.14-2-powerpc: Installation fails #345067: [powerpc] ide-generic is not built on powerpc, yaird tries to include it and fails Both relate to ide-generic. Difference between yaird and initramfs-tools in regards to this issue is that yaird has builtin probing while initramfs-tools rely on udev for extracting kernels own logic and/or implement workarounds. What has that to do with anything ? The question was should yaird not be made the default and i answered that this is probably not a good idea because the DD maintainer (you) doesn't seem able to fix bugs without consulting his upstream and that said upstream is MIA. Sven Luther -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#349354: why not yaird by default?
This one time, at band camp, friendly Sven Luther said: On Mon, Feb 27, 2006 at 03:43:17PM +0100, Jonas Smedegaard wrote: On Mon, 27 Feb 2006 15:14:02 +0100 Sven Luther [EMAIL PROTECTED] wrote: #343427: linux-image-2.6.14-2-powerpc: Installation fails #345067: [powerpc] ide-generic is not built on powerpc, yaird tries to include it and fails Both relate to ide-generic. Difference between yaird and initramfs-tools in regards to this issue is that yaird has builtin probing while initramfs-tools rely on udev for extracting kernels own logic and/or implement workarounds. What has that to do with anything ? Since both bugs are arguably kernel bugs (some modules on some platforms can't work without also loading ide-generic, but the kernel provides no mechanism to find that out), I think it has rather a lot to do with the issue at hand. The question was should yaird not be made the default and i answered that this is probably not a good idea because the DD maintainer (you) doesn't seem able to fix bugs without consulting his upstream and that said upstream is MIA. An MIA upstream is indeed a serious problem. A maintainer being unwilling to accept a bad hack to work around brokenness elsewhere is less of an issue, at least IMHO. Ever so friendly, -- - | ,''`.Stephen Gran | | : :' :[EMAIL PROTECTED] | | `. `'Debian user, admin, and developer | |`- http://www.debian.org | - signature.asc Description: Digital signature
Bug#349354: why not yaird by default?
Is there any drawback to just switching the default dependency to be on yaird instead of initramfs-tools? -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#349354: why not yaird by default?
On Fri, Feb 24, 2006 at 04:24:55PM -0500, Mike O'Connor wrote: Is there any drawback to just switching the default dependency to be on yaird instead of initramfs-tools? Yes. You can't install yaird on upgrade from a 2.4 kernel. And currently, you can't install initramfs-tools on upgrade from a 2.6.8 kernel. In each case there are workarounds, but we really want to be able to break this dependency loop. -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. [EMAIL PROTECTED] http://www.debian.org/ signature.asc Description: Digital signature
Bug#349354: why not yaird by default?
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Fri, 24 Feb 2006 14:09:29 -0800 Steve Langasek [EMAIL PROTECTED] wrote: On Fri, Feb 24, 2006 at 04:24:55PM -0500, Mike O'Connor wrote: Is there any drawback to just switching the default dependency to be on yaird instead of initramfs-tools? Yes. You can't install yaird on upgrade from a 2.4 kernel. And currently, you can't install initramfs-tools on upgrade from a 2.6.8 kernel. In each case there are workarounds, but we really want to be able to break this dependency loop. More details on the limitations of each ramdisk tools (and other comparative info) is on the Debian wiki[1]. Please help fill in the gaps at that page, if you know of something that is not on that page, or is marked wrongly or as Unknown, needs to be tested :-) - Jonas [1] http://wiki.debian.org/InitrdReplacementOptions - -- * Jonas Smedegaard - idealist og Internet-arkitekt * Tlf.: +45 40843136 Website: http://dr.jones.dk/ - Enden er nær: http://www.shibumi.org/eoti.htm -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.2 (GNU/Linux) iD8DBQFD/4mmn7DbMsAkQLgRApfUAKCPXB5zw/ehcqpIwfFD0L0sMA1N6QCgm10E bAqa0sqanAZ8+x65WcxebQU= =SOQy -END PGP SIGNATURE-