Same system as before. This time I burned a heap of CDs to try a CD install. Fairly uneventful (I knew what to do about the disk partitioning by now), and the display is much better with "resolution=1280x800"
Actually booting was a problem. The system atorebooted about immediately unpacking the kernel.
Finding an SMP kernel as the only one was a surprise.
No matter, boot CD1 in rescue mode. Eh? No kernel on CD1? Damn, never mind, I'll install off the LAN.
Why did that fail? Did that duff signature matter? No? Why's that scriptlet failing? --noscripts doesn't work? some mumble about recon<something> rang a bell. How to I disable selinue? The likely commands get errors.
Reboot, selinux=0 and all's well, I can install the kernel and the new kernel boots.
I was writing this as the system booted, so I don't know what happened to firstboot; when I turned back to the laptop it was at the GDM login.
I've not had a chance to create a user account.
Okay, I logged in to create a user account. Quick question, why's "terminal" not in the "Accessories" list where a Windows XP user would expect to find it?
I create /home/userd (meaning /home/users) then used to GUI to create "summer" and O iverrode the UID to 1000 and and home to /home/users/summer. There was no error reported (recall /home/users didn't exist) and summer was homeless.
Previously, when I've create a user accound on FC/EL systems I've used the CLI tools (useradd|adduser and passwd). When UID 1000 is used, as here, the next UID used it 1001, not 500. I prefer to continue from 1000 (compatible with other distros, those that run from 1000 might break if a sysadmin forced 500).
Something needs to be done about wireless. SuSE 10.0 and Ubuntu can both configure my Atheros wireless, Mandriva knw the wiresless there and wanted to ser up ndiswrapper for me: I couldn't coz I don't know where in Windows the driver is, but at least it tried.
Fedora Core 5 beta 1 doesn't even know the thing is there, when I try to configure "wireless" the only choice I have is "other wireless" and that leads me to a selection of drivers including NE1000 and many other (not) wireless devices.
At the equivalent point in SuSE, Yast recognised the device and asked me to configure it, installed the driver and set it up. I don't recall exactly what Ubuntu did, but the driver was ready installed and it was just a matter of configuring it, just like any network device.
There's also the question of the infernal modem. Again, SuSE lead me to it, installed the driver and configured it.
I wish to use the Atheros driver; I understand RH won't ship the HAL, but it needs to make it easy for users to implement their right to choose to use it.
Same for the internal modem. Ordinary users cannot be expected to have the skills or desire to hunt round the Internet for the driver.
I could argue FC5 and EL4 are ready for the desktop, but they certainly are not ready for the laptop.
Plugging in a PCMCIA Orinoco 11b wireless card resulted in two! interfaces, neither the eth1 I expected (and get on FC3). Instead, I have wlan0 and wifi0. Further they have the default essid "test" and I cannot set it to "" or "any." Nore can I set the WEP key: # iwconfig wlan0 essid anaglorious key s:secre Error for wireless request "Set Encode" (8B2A) : SET failed on device wlan0 ; Invalid argument.
There are also lots of these messages: search_node dbd094b4 start_node dbd094b4 return_node 00000000 ACPI-0508: *** Error: Method execution failed [_SB_.BAT1._BST] (Node dbd098d4), AE_NOT_FOUND ACPI-0339: *** Error: Looking up [Z004] in namespace, AE_NOT_FOUND search_node dbd094b4 start_node dbd094b4 return_node 00000000 ACPI-0508: *** Error: Method execution failed [_SB_.BAT1._BST] (Node dbd098d4), AE_NOT_FOUND
fwiw Ununtu and SuSE have them too. Google showed some discussion on lkml.
On Mon, Nov 28, 2005 at 03:05:39PM +0800, John Summerfied wrote:
Same system as before. This time I burned a heap of CDs to try a CD install. Fairly uneventful (I knew what to do about the disk partitioning by now), and the display is much better with "resolution=1280x800"
Actually booting was a problem. The system atorebooted about immediately unpacking the kernel.
Finding an SMP kernel as the only one was a surprise.
As the fine release notes mentioned, booting with mem=nopentium 'fixes' this.
Reboot, selinux=0 and all's well, I can install the kernel and the new kernel boots.
Odd that selinux stopped the reboot bug. Anyways, as you've found out, the latest kernel in rawhide fixes this bug properly, so it should 'just boot'.
I wish to use the Atheros driver; I understand RH won't ship the HAL, but it needs to make it easy for users to implement their right to choose to use it.
Same for the internal modem. Ordinary users cannot be expected to have the skills or desire to hunt round the Internet for the driver.
Those other distros you mention ship "not merged upstream" drivers. Fedora doesn't do that. Look at the mess we had with the ipw2xxx drivers. Those were just two drivers we were carrying before they went upstream, and it was a nightmare trying to keep them in sync with upstream. Now imagine the same problem with a few dozen drivers ? The answer is "ask these projects to get their work upstream".
Plugging in a PCMCIA Orinoco 11b wireless card resulted in two! interfaces, neither the eth1 I expected (and get on FC3). Instead, I have wlan0 and wifi0.
No idea on this one. The orinoco driver should only create ethX interfaces. What drivers are in use ?
There are also lots of these messages: search_node dbd094b4 start_node dbd094b4 return_node 00000000 ACPI-0508: *** Error: Method execution failed [_SB_.BAT1._BST] (Node dbd098d4), AE_NOT_FOUND ACPI-0339: *** Error: Looking up [Z004] in namespace, AE_NOT_FOUND search_node dbd094b4 start_node dbd094b4 return_node 00000000 ACPI-0508: *** Error: Method execution failed [_SB_.BAT1._BST] (Node dbd098d4), AE_NOT_FOUND
fwiw Ununtu and SuSE have them too. Google showed some discussion on lkml.
ACPI debug spew de jour ;-) That looks familiar, I think I've seen at least 1-2 instances of that in bugzilla. Battery reporting seems to have a few issues in 2.6.14/15
Dave
Dave Jones wrote:
On Mon, Nov 28, 2005 at 03:05:39PM +0800, John Summerfied wrote:
Same system as before. This time I burned a heap of CDs to try a CD install. Fairly uneventful (I knew what to do about the disk partitioning by now), and the display is much better with "resolution=1280x800"
Actually booting was a problem. The system atorebooted about immediately unpacking the kernel.
Finding an SMP kernel as the only one was a surprise.
As the fine release notes mentioned, booting with mem=nopentium 'fixes' this.
Reboot, selinux=0 and all's well, I can install the kernel and the new kernel boots.
Odd that selinux stopped the reboot bug. Anyways, as you've found out, the latest kernel in rawhide fixes this bug properly, so it should 'just boot'.
I wish to use the Atheros driver; I understand RH won't ship the HAL, but it needs to make it easy for users to implement their right to choose to use it.
Same for the internal modem. Ordinary users cannot be expected to have the skills or desire to hunt round the Internet for the driver.
Those other distros you mention ship "not merged upstream" drivers. Fedora doesn't do that. Look at the mess we had with the ipw2xxx drivers. Those were just two drivers we were carrying before they went upstream, and it was a nightmare trying to keep them in sync with upstream. Now imagine the same problem with a few dozen drivers ? The answer is "ask these projects to get their work upstream".
I probably wasn't using wireless then, maybe wasn't using FC either: I used Debian for some time, and when I started using wireless it was with prism54 in or after 2.6.8 and only had to worry about firmware.
I do recall comparing the number of lines of Linux kernel with the number of lines of RH patches back around RHL 7.3. Are you saying RH has changed its ways?
I've taken the easy way out: I reinstalled SUSE 10.0.*
Plugging in a PCMCIA Orinoco 11b wireless card resulted in two! interfaces, neither the eth1 I expected (and get on FC3). Instead, I have wlan0 and wifi0.
No idea on this one. The orinoco driver should only create ethX interfaces. What drivers are in use ?
I'm certain it was orinoco: did take a look to see whether I could see anything else odd.
There are also lots of these messages: search_node dbd094b4 start_node dbd094b4 return_node 00000000 ACPI-0508: *** Error: Method execution failed [_SB_.BAT1._BST] (Node dbd098d4), AE_NOT_FOUND ACPI-0339: *** Error: Looking up [Z004] in namespace, AE_NOT_FOUND search_node dbd094b4 start_node dbd094b4 return_node 00000000 ACPI-0508: *** Error: Method execution failed [_SB_.BAT1._BST] (Node dbd098d4), AE_NOT_FOUND
fwiw Ununtu and SuSE have them too. Google showed some discussion on lkml.
ACPI debug spew de jour ;-) That looks familiar, I think I've seen at least 1-2 instances of that in bugzilla. Battery reporting seems to have a few issues in 2.6.14/15
Google suggest it was well-enough known that there probably wasn't any point in reporting it. OTOH I thought to ensure it's a known problem here too.
* I have to reorganise the drive, there is currentlly space the partition tables won't let me use.
On Tue, Nov 29, 2005 at 07:40:42AM +0800, John Summerfied wrote:
I do recall comparing the number of lines of Linux kernel with the number of lines of RH patches back around RHL 7.3. Are you saying RH has changed its ways?
Fedora has had the goal of being as close to upstream as possible since day one. We still carry some patches, but "extra feature/driver" is something we try to avoid. The disaster we had with supporting ipw over the last few releases is a good example why.
You either pick an old kernel, and stick with it for the lifetime of a release, or you get the option of rebasing to a newer upstream kernel as/when they become available. Doing the former means any add-on drivers you carry won't break, but it does mean you have the nightmare job of backporting fixes from the later upstream releases, which given the current upstream rate of ~4-5000 changes per release (bi-monthly, though moves are afoot to try and increase the release rate), really isn't feasible.
So we rebase to each new point release. The amount of effort involved is trivial. The daily rediffing that happens in rawhide hits a conflict that needs resolving once every so often, and is 2-3 minutes work.
There are currently 90 patches applied in todays rawhide kernel. A bunch of them will never go upstream (exec-shield for eg, though some bits of it did get merged), a bunch of others have been merged in upstream maintainers -git trees, but haven't found their way to Linus' tree yet. There are a bunch of other patches that could go upstream, and that process typically goes..
- make patch for bug in bugzilla - send patch upstream, and commit to cvs - when daily rebase fails due to 'already applied patch', drop patch
Most of those are in the second stage, some may need retransmitting.
But on the whole, things are very different from RHL7 days. looking at the number of lines of patches, Fedora has a lot less change as what RHL7 did.
RHL7.3 - 210272 lines rawhide - 43861 lines.
That's still a lot higher than I'd like, but we're getting there.
Dave
Dave Jones wrote:
On Tue, Nov 29, 2005 at 07:40:42AM +0800, John Summerfied wrote:
I do recall comparing the number of lines of Linux kernel with the number of lines of RH patches back around RHL 7.3. Are you saying RH has changed its ways?
Fedora has had the goal of being as close to upstream as possible since day one. We still carry some patches, but "extra feature/driver" is something we try to avoid. The disaster we had with supporting ipw over the last few releases is a good example why.
Would it then be possible to include a tool that facilitates users' fetching and building some not-ready-for-prime-time drivers such as the madwifi drivers, drivers for common sofmodems? It might not need more than an additional yum repository and some commitment with others to work together to keep drivers synchronised and a (maybe dummy) package for users to install that "requires" matching kernel and drivers packages so yum and its friends and competitors can ensure, when users using both kernel and brand-X-wireless-driver don't get a new kernel unless they can also get the matchiing brand-X-wireless-driver.
ATM support for laptops in the various distros is very variable.
Ubuntu (breezy) installed the madwifi driver (and I know handles some other 11g wireless too) out of the box. I didn't notice that it did anything for my softmodem. Ubuntu's stated aim for the next release is pretty much what OS X does: basically, once it's configured (and that's easy), the Mac checks for APs it knows and attaches to one.
SUSE 10 installed the drivers for both fine, but I suspect it will fall down in moving between home and office (different APs).
Mandriva offered to install ndiswrapper, but neither I nor M knew where in my XP system to find the driver.
FC does nothing.