Re: [arch-general] [arch-dev-public] [signoff] initscripts-2011.04.1-2
Am Fri, 29 Apr 2011 02:33:57 +0200 schrieb Tom Gundersen <teg@jklm.no>:
* We now strongly discourage the use of HWCLOCK="localtime", as this may lead to several known and unfixable bugs
This can lead to problems on multi boot systems on which Arch Linux is installed parallel to Windows. At least older Windows versions can't handle hardware clocks which are set to UTC. So if there are issues with HWCLOCK="localtime" these should be fixed instead. Heiko
On Fri, Apr 29, 2011 at 8:43 AM, Heiko Baums <lists@baums-on-web.de> wrote:
Am Fri, 29 Apr 2011 02:33:57 +0200 schrieb Tom Gundersen <teg@jklm.no>:
* We now strongly discourage the use of HWCLOCK="localtime", as this may lead to several known and unfixable bugs
This can lead to problems on multi boot systems on which Arch Linux is installed parallel to Windows. At least older Windows versions can't handle hardware clocks which are set to UTC.
So if there are issues with HWCLOCK="localtime" these should be fixed instead.
Heiko
I was under the impression that windows versions from Windows XP onwards had a registry key which enabled them to use UTC time.
Am Fri, 29 Apr 2011 09:01:23 +0800 schrieb Oon-Ee Ng <ngoonee.talk@gmail.com>:
I was under the impression that windows versions from Windows XP onwards had a registry key which enabled them to use UTC time.
Well, I'm not using Windows anymore for many years. So I can be wrong, but I didn't know such a registry key, yet. Now I found one blog and one forum which point to such a registry key, but you still have to use a registry editor or tweaker on Windows. So it's not a common option in Windows, and I guess it's pretty unknown. I'd suggest to better assume that Windows users usual have their hardware clock set to localtime. Heiko
On Fri, Apr 29, 2011 at 9:19 AM, Heiko Baums <lists@baums-on-web.de> wrote:
Am Fri, 29 Apr 2011 09:01:23 +0800 schrieb Oon-Ee Ng <ngoonee.talk@gmail.com>:
I was under the impression that windows versions from Windows XP onwards had a registry key which enabled them to use UTC time.
Well, I'm not using Windows anymore for many years. So I can be wrong, but I didn't know such a registry key, yet.
Now I found one blog and one forum which point to such a registry key, but you still have to use a registry editor or tweaker on Windows. So it's not a common option in Windows, and I guess it's pretty unknown. I'd suggest to better assume that Windows users usual have their hardware clock set to localtime.
Conversely, perhaps we can assume that Windows users who are able to install Arch Linux (and dual-boot, no less) are able to edit a registry key =). I have two dual-boot systems, I use UTC on one and localtime on the other, based on what was convenient at the time. If there's bugs with localtime switching to UTC isn't hard, and is probably easier than fixing bugs for an easily-worked-around issue.
Am Fri, 29 Apr 2011 09:38:51 +0800 schrieb Oon-Ee Ng <ngoonee.talk@gmail.com>:
Conversely, perhaps we can assume that Windows users who are able to install Arch Linux (and dual-boot, no less) are able to edit a registry key =).
I have two dual-boot systems, I use UTC on one and localtime on the other, based on what was convenient at the time. If there's bugs with localtime switching to UTC isn't hard, and is probably easier than fixing bugs for an easily-worked-around issue.
Are those bugs Arch related or do they affect other distros, too? If they are just Arch related then I would not object. But if they affect other distros, too, then I disagree, because there are several distros which can easily be installed, also as dual-boot systems, by beginners who aren't very familiar with the Windows registry. Heiko
On Fri, Apr 29, 2011 at 9:52 AM, Heiko Baums <lists@baums-on-web.de> wrote:
Am Fri, 29 Apr 2011 09:38:51 +0800 schrieb Oon-Ee Ng <ngoonee.talk@gmail.com>:
Conversely, perhaps we can assume that Windows users who are able to install Arch Linux (and dual-boot, no less) are able to edit a registry key =).
I have two dual-boot systems, I use UTC on one and localtime on the other, based on what was convenient at the time. If there's bugs with localtime switching to UTC isn't hard, and is probably easier than fixing bugs for an easily-worked-around issue.
Are those bugs Arch related or do they affect other distros, too?
If they are just Arch related then I would not object. But if they affect other distros, too, then I disagree, because there are several distros which can easily be installed, also as dual-boot systems, by beginners who aren't very familiar with the Windows registry.
Isn't initscripts an Arch-specific project?
Am Fri, 29 Apr 2011 10:04:30 +0800 schrieb Oon-Ee Ng <ngoonee.talk@gmail.com>:
Isn't initscripts an Arch-specific project?
I don't know if those problems are just the initscripts or if localtime shall just be ignored by the initscripts due to upstream issues. I even don't know the issues. Well, I don't care that much, because I'm not using Windows anymore and have set my hardware clock to UTC anyway. I'm just giving my objections. Heiko
On Fri, Apr 29, 2011 at 03:19:04AM +0200, Heiko Baums wrote:
Am Fri, 29 Apr 2011 09:01:23 +0800 schrieb Oon-Ee Ng <ngoonee.talk@gmail.com>:
I was under the impression that windows versions from Windows XP onwards had a registry key which enabled them to use UTC time.
Well, I'm not using Windows anymore for many years. So I can be wrong, but I didn't know such a registry key, yet.
Now I found one blog and one forum which point to such a registry key, but you still have to use a registry editor or tweaker on Windows. So it's not a common option in Windows, and I guess it's pretty unknown. I'd suggest to better assume that Windows users usual have their hardware clock set to localtime.
Well, i sometimes use windows xp, though only for gaming. I didn't know that windows xp could be set to use UTC, even through a registry entry. After reading through this thread, i found info on how to enable UTC on windows. But there have been some bugs reported as well. So, if we indeed need to discourage use of localtime in rc.conf, firstly there should be a news entry about this. And if possible, post-installation, there should be instructions on how to enable UTC on windows.
On Thu, Apr 28, 2011 at 9:10 PM, gt <codered12@gmail.com> wrote:
On Fri, Apr 29, 2011 at 03:19:04AM +0200, Heiko Baums wrote:
Am Fri, 29 Apr 2011 09:01:23 +0800 schrieb Oon-Ee Ng <ngoonee.talk@gmail.com>:
I was under the impression that windows versions from Windows XP onwards had a registry key which enabled them to use UTC time.
Well, I'm not using Windows anymore for many years. So I can be wrong, but I didn't know such a registry key, yet.
Now I found one blog and one forum which point to such a registry key, but you still have to use a registry editor or tweaker on Windows. So it's not a common option in Windows, and I guess it's pretty unknown. I'd suggest to better assume that Windows users usual have their hardware clock set to localtime.
Well, i sometimes use windows xp, though only for gaming. I didn't know that windows xp could be set to use UTC, even through a registry entry.
After reading through this thread, i found info on how to enable UTC on windows. But there have been some bugs reported as well. So, if we indeed need to discourage use of localtime in rc.conf, firstly there should be a news entry about this. And if possible, post-installation, there should be instructions on how to enable UTC on windows.
a [brief] post-installation message seems a good compromise to me -- maybe even the registry file itself too -- but in the absence of both this is still a good change imo. setting the hardware clock to localtime is 100% broken and the links provided in the original announcement explain the ramifications clearly ... at least i thought there were links somewhere ... wtf, where did i see them? it was an arch-related mail because i read the link ... at any rate, a 5 second google search turned up: https://help.ubuntu.com/community/UbuntuTime?action=AttachFile&do=get&target=WindowsTimeFixUTC.reg http://kb.norsetech.net/set-windows-clock-to-utc-time/ download. double-click. done :-) C Anthony
Am Thu, 28 Apr 2011 22:22:39 -0500 schrieb C Anthony Risinger <anthony@extof.me>:
a [brief] post-installation message seems a good compromise to me -- maybe even the registry file itself too -- but in the absence of both this is still a good change imo.
This should not only be a brief post-installation message. An announcement in the Latest News on the homepage is important, too.
https://help.ubuntu.com/community/UbuntuTime?action=AttachFile&do=get&target=WindowsTimeFixUTC.reg http://kb.norsetech.net/set-windows-clock-to-utc-time/
download. double-click. done :-)
Those links, the instructions and/or the reg file should be given, too, in this annoucement. Heiko
On 04/28/2011 11:22 PM, C Anthony Risinger wrote:
setting the hardware clock to localtime is 100% broken and the links provided in the original announcement explain the ramifications clearly ... at least i thought there were links somewhere ... wtf, where did i see them? it was an arch-related mail because i read the link ...
Can someone pls clarify what exactly is broken about localtime? I've been using it for years without any (noticable) issue. I'll be happy to switch over if I need to, but I'd like to understand the problem(s) first. DR
On Fri, Apr 29, 2011 at 8:41 PM, David Rosenstrauch <darose@darose.net> wrote:
Can someone pls clarify what exactly is broken about localtime? I've been using it for years without any (noticable) issue. I'll be happy to switch over if I need to, but I'd like to understand the problem(s) first.
There are some implementational problems and some conceptual ones. Please have a look at the link I gave to the systemd page for a brief description as well as <http://www.cl.cam.ac.uk/~mgk25/mswish/ut-rtc.html> for the very interested. In short (sorry for any mistakes or omissions): In linux we have (at least) two clocks, the hardware clock and the system time. The system time is what matters when the computer is running and this is always in UTC, the hardware clock is what remembers the time for us when the computer is switched off and this is what we are discussing now. It may be in localtime or in UTC. When we turn the computer on we have to instantiate the system time from the hardware clock. If the hardware clock is in UTC this is straightforward, but if it is in localtime it is impossible. Well, it is impossible to get it right all of the time, but it almost always works out ok (that's why most people will never notice). There are (at least) two reasons for the problems and they both have to do with daylight saving time: 1) as I said, when the computer starts we need to take the time from the hardware clock and adjust for the timezone to get the time in UTC. However, imagine that your hardware clock tells you it is half past two in the morning on the day the clock should be adjusted back by one hour. What time is it in UTC? There is no way to know (it depends on whether the clock was adjusted back half an hour ago, or if it should be adjusted back in half an hour). This problem occurs in a one hour window every year. 2) if you are dual-booting, who should correct for DST? Usually both OS'es will, and then we are off by one hour. If you decide that only one should do it then we are off by one hour if we start the wrong one first, etc. This problem occurs twice every year. HTH, Tom
On Fri 29 Apr 2011 23:00 +0200, Tom Gundersen wrote:
On Fri, Apr 29, 2011 at 8:41 PM, David Rosenstrauch <darose@darose.net> wrote:
Can someone pls clarify what exactly is broken about localtime? I've been using it for years without any (noticable) issue. I'll be happy to switch over if I need to, but I'd like to understand the problem(s) first.
There are some implementational problems and some conceptual ones. Please have a look at the link I gave to the systemd page for a brief description as well as <http://www.cl.cam.ac.uk/~mgk25/mswish/ut-rtc.html> for the very interested.
Sounds like a hardware problem then. Maybe the bios should tell the OS if the clock is in localtime or UTC. ;)
On 30-04-2011 00:51, Loui Chang wrote:
Sounds like a hardware problem then. Maybe the bios should tell the OS if the clock is in localtime or UTC. ;)
Not really, the real time clock is a clock, it is supposed to keep the time always moving forward in the most accurate way possible, ideally it would keep in sync with a worldwide accepted time reference, UTC.
From the links presented so far and from my point of view the problem is with the localtime concept itself, it was broken from the start, mucking about with the time source is a bad idea and I can understand perfectly why it's use is discouraged.
-- Mauro Santos
Am 29.04.2011 02:43, schrieb Heiko Baums:
Am Fri, 29 Apr 2011 02:33:57 +0200 schrieb Tom Gundersen <teg@jklm.no>:
* We now strongly discourage the use of HWCLOCK="localtime", as this may lead to several known and unfixable bugs
This can lead to problems on multi boot systems on which Arch Linux is installed parallel to Windows. At least older Windows versions can't handle hardware clocks which are set to UTC.
So if there are issues with HWCLOCK="localtime" these should be fixed instead.
Frankly, I don't care. The problems with localtime are so numerous and subtle, handling them all is a waste of time - and it gets worst when you change your time zones often. Using UTC is the logical approach and works without bugs. Every recent operating system I know can handle UTC, even Windows Vista (bugs before SP1, works with SP1 or later) and Windows 7, and Windows XP is so old, it shouldn't be used anyway.
Am Fri, 29 Apr 2011 11:02:22 +0200 schrieb Thomas Bächler <thomas@archlinux.org>:
Every recent operating system I know can handle UTC, even Windows Vista (bugs before SP1, works with SP1 or later) and Windows 7, and Windows XP is so old, it shouldn't be used anyway.
But there's a problem with Windows. Updating Windows is not as easy for everybody as Linux is. There's a substantial buying resistance. So most Windows users stay with their Windows version they once bought their whole life, at least for the whole lifetime of their computer. And as far as I know Windows XP shall be much better than Windows Vista. Heiko
On 04/29/2011 06:46 AM, Heiko Baums wrote:
Am Fri, 29 Apr 2011 11:02:22 +0200 schrieb Thomas Bächler<thomas@archlinux.org>:
Every recent operating system I know can handle UTC, even Windows Vista (bugs before SP1, works with SP1 or later) and Windows 7, and Windows XP is so old, it shouldn't be used anyway. But there's a problem with Windows. Updating Windows is not as easy for everybody as Linux is. There's a substantial buying resistance. So most Windows users stay with their Windows version they once bought their whole life, at least for the whole lifetime of their computer.
And as far as I know Windows XP shall be much better than Windows Vista.
Heiko I'm with Thomas here. If it bugs you that much submit some patches. It seems as if you always have an opinion, but rarely have a solution and/or patches.
Folks, don't forget, arch isn't a democracy. The developers do what they want. pyther
Am Fri, 29 Apr 2011 07:07:58 -0400 schrieb Matthew Gyurgyik <pyther@pyther.net>:
I'm with Thomas here. If it bugs you that much submit some patches. It seems as if you always have an opinion, but rarely have a solution and/or patches.
And I say my opinion, at least if I see that there could be regressions or serious issues. If it's just a misunderstanding due to too little informations given with an announcement it can easily be cleared up by explaining it a bit more detailed. I'm not a developer, so this may be one of the reasons for not giving patches or solutions so many times. If I have one, I usually give it, and I did it quite a few times already.
Folks, don't forget, arch isn't a democracy. The developers do what they want.
Then the developers should just develop for themselves without releasing it publically or without creating a community around it, if they don't want to listen to the users and their opinions. And/or they should just say on the homepage: "You may use it, but if it fails you're on your own." But I doubt that this is the opinion of most of the developers. Heiko
2011/4/29 Heiko Baums <lists@baums-on-web.de>:
Am Fri, 29 Apr 2011 07:07:58 -0400 schrieb Matthew Gyurgyik <pyther@pyther.net>:
I'm with Thomas here. If it bugs you that much submit some patches. It seems as if you always have an opinion, but rarely have a solution and/or patches.
And I say my opinion, at least if I see that there could be regressions or serious issues. If it's just a misunderstanding due to too little informations given with an announcement it can easily be cleared up by explaining it a bit more detailed.
I'm not a developer, so this may be one of the reasons for not giving patches or solutions so many times. If I have one, I usually give it, and I did it quite a few times already.
Folks, don't forget, arch isn't a democracy. The developers do what they want.
Then the developers should just develop for themselves without releasing it publically or without creating a community around it, if they don't want to listen to the users and their opinions. And/or they should just say on the homepage: "You may use it, but if it fails you're on your own." But I doubt that this is the opinion of most of the developers.
Heiko
Ok are you freaking joking on me? Why we have to worry about other OS? if they do the wrong things, then report it to them, we don't have to eat a hundreds of bugs, just because you like a crappy os just for game. That said, good job Tom! -- Angel Velásquez angvp @ irc.freenode.net Arch Linux Developer / Trusted User Linux Counter: #359909 http://www.angvp.com
Am Fri, 29 Apr 2011 11:18:16 -0300 schrieb Ángel Velásquez <angvp@archlinux.org>:
Ok are you freaking joking on me? Why we have to worry about other OS? if they do the wrong things, then report it to them, we don't have to eat a hundreds of bugs, just because you like a crappy os just for game.
If you had read what I have written, then you'd know that I personally don't care much about that as I'm not using another OS anymore for a long time. But there are a lot of people who are using another OS and who need another OS in a dual-boot system, e.g. people who first want to try out and get to know Linux before switching completely. That was how I switched to Linux. And you'd know that I didn't know that Windows can be set to use UTC. And I bet that most Windows users don't know this. And, even if I don't like to say that, it's the question if Windows is broken, if it doesn't support UTC officially in an easy way, or Linux if localtime is broken, or if both are broken. So it should not only be fixed in Windows but also in Linux. If this is not possible for whatever reason then this is not possible, but it can be easily explained in a short note instead of starting such a long discussion or saying such a nonsense like: "Folks, don't forget, arch isn't a democracy. The developers do what they want." This hasn't anything to do with this topic and really doesn't help. Heiko
Am 29.04.2011 15:36, schrieb Heiko Baums:
And I say my opinion, at least if I see that there could be regressions or serious issues.
If I remember it right, the only changes were the default value and comment in rc.conf.
Am Fri, 29 Apr 2011 17:11:22 +0200 schrieb Thomas Bächler <thomas@archlinux.org>:
If I remember it right, the only changes were the default value and comment in rc.conf.
Maybe, and it's alright. But as it wasn't released, yet, I just wanted to note that a lot of people could get problems if localtime isn't supported anymore. And I wanted to note it before it is too late. ;-) Nevertheless an announcement should indeed be written in the Latest News and probably in the Beginner's Guide with a link to the reg file or a short explanation for the other OS. Heiko
2011/4/29 Heiko Baums <lists@baums-on-web.de>:
Am Fri, 29 Apr 2011 17:11:22 +0200 schrieb Thomas Bächler <thomas@archlinux.org>:
If I remember it right, the only changes were the default value and comment in rc.conf.
Maybe, and it's alright. But as it wasn't released, yet, I just wanted to note that a lot of people could get problems if localtime isn't supported anymore. And I wanted to note it before it is too late. ;-)
Nevertheless an announcement should indeed be written in the Latest News and probably in the Beginner's Guide with a link to the reg file or a short explanation for the other OS.
Ok stop thinking in other OS we can't assume or try to guess all the posible cases from how you installed Arch. Why we have to put this silly info for users of XX OS when they aren't aware of the existence of other OS and don't care if they make stuff uncompatible?. This is silly, and as a user, not developer, talking by myself not for the distro, I particularly think, that we should avoid to stop development because uncompatibilities with XX OS, our goal is Arch, not the rest of the OS.
Heiko
-- Angel Velásquez angvp @ irc.freenode.net Arch Linux Developer / Trusted User Linux Counter: #359909 http://www.angvp.com
On Fri, Apr 29, 2011 at 10:16 PM, Heiko Baums <lists@baums-on-web.de> wrote:
[...] if localtime isn't supported anymore.
Sorry to be repeating myself: we are not considering dropping support for localtime. We are just pointing out that it is a bad idea to use it (it has always been a bad idea, nothing has changed, we are just pointing it out, as it is not obvious).
Nevertheless an announcement should indeed be written in the Latest News
It will be (my first email contained a draft of that).
and probably in the Beginner's Guide with a link to the reg file or a short explanation for the other OS.
Sounds like a good idea, the link I gave to the systemd page would be a good starting point. Contributions welcome! -t
On Fri, Apr 29, 2011 at 5:38 PM, Tom Gundersen <teg@jklm.no> wrote:
On Fri, Apr 29, 2011 at 10:16 PM, Heiko Baums <lists@baums-on-web.de> wrote:
[...] if localtime isn't supported anymore.
Sorry to be repeating myself: we are not considering dropping support for localtime.
Yeah, you are right. At first sight (or maybe an oversight...) I've got the impression of it being unsupported from now on. But rereading the thread made that clear. Maybe that could be made more explicit. -- A: Because it obfuscates the reading. Q: Why is top posting so bad? ------------------------------------------- Denis A. Altoe Falqueto Linux user #524555 -------------------------------------------
On Fri, Apr 29, 2011 at 10:44 PM, Denis A. Altoé Falqueto <denisfalqueto@gmail.com> wrote:
Yeah, you are right. At first sight (or maybe an oversight...) I've got the impression of it being unsupported from now on. But rereading the thread made that clear. Maybe that could be made more explicit.
Yes, it appears this should be made more clear. I'll write: "* We now strongly discourage the use of HWCLOCK="localtime", as this may lead to several known and unfixable bugs. However, there are no plans to drop support for localtime."
On Fri, Apr 29, 2011 at 5:11 PM, Thomas Bächler <thomas@archlinux.org> wrote:
Am 29.04.2011 15:36, schrieb Heiko Baums:
And I say my opinion, at least if I see that there could be regressions or serious issues.
If I remember it right, the only changes were the default value and comment in rc.conf.
Correct. -t
Hi all, I'll try to answer all the comments in one go: On Fri, Apr 29, 2011 at 2:43 AM, Heiko Baums <lists@baums-on-web.de> wrote:
Am Fri, 29 Apr 2011 02:33:57 +0200 schrieb Tom Gundersen <teg@jklm.no>:
* We now strongly discourage the use of HWCLOCK="localtime", as this may lead to several known and unfixable bugs
This can lead to problems on multi boot systems on which Arch Linux is installed parallel to Windows. At least older Windows versions can't handle hardware clocks which are set to UTC.
Windows can be configured to use UTC. If this is not possible, we still support localtime, but it will cause weird bugs. The only change we made was to change the default and to add a comment to rc.conf. Nothing will change on existing systems.
So if there are issues with HWCLOCK="localtime" these should be fixed instead.
I really meant what I said: The problems cannot be fixed, they are conceptual, not implementation specific. I wrote a summary of why systemd does not support localtime, the same problems apply to initscripts (but we have just been ignoring them): <https://wiki.archlinux.org/index.php/Systemd#Why_does_systemd_not_support_the_RTC_being_in_localtime.3F>. I don't think any additional announcement should be necessary, except for the above note on the front page and the comment in rc.conf. It would of course be very helpful if people could update any documentation they find that mentions localtime/UTC. Hope that answered all the questions, Tom
participants (12)
-
C Anthony Risinger
-
David Rosenstrauch
-
Denis A. Altoé Falqueto
-
gt
-
Heiko Baums
-
Loui Chang
-
Matthew Gyurgyik
-
Mauro Santos
-
Oon-Ee Ng
-
Thomas Bächler
-
Tom Gundersen
-
Ángel Velásquez