Bugs in new kernels

[Lloyd Hayes]

Well, I got fed up with the bugs in Fedora Core 4 and put Fedora Core 3
back on my computer.

FC3 came out a year ago. One of the first things that I noticed was that
some of newer programs did not work well with the Selinx policies.
Updating the software took sever hours. (All night)

This morning, after all of the downloading, I was almost required to
install the new kernel first.

After that, the bugs of FC4 showed up. They had not been there from the
original kernel on the installation CD. They showed up after installing
the new kernel. It appears that the bugs associated with FC4 are related
to the new kernels being used. The same bugs show up in FC3 when the
latest kernel is applied.

[Alexander Dalloz]

Am Mo, den 01.08.2005 schrieb Lloyd Hayes um 0:04:
> Well, I got fed up with the bugs in Fedora Core 4 and put Fedora Core 3
> back on my computer.
>
> FC3 came out a year ago. One of the first things that I noticed was that
> some of newer programs did not work well with the Selinx policies.
> Updating the software took sever hours. (All night)
>
> This morning, after all of the downloading, I was almost required to
> install the new kernel first.
>
> After that, the bugs of FC4 showed up. They had not been there from the
> original kernel on the installation CD. They showed up after installing
> the new kernel. It appears that the bugs associated with FC4 are related
> to the new kernels being used. The same bugs show up in FC3 when the
> latest kernel is applied.

What do you want to tell us with that message? Do you in parallel ticket
bugs you found in bugzilla with proper detailed information? Then please
let us know the ticket numbers.
Sad to say, else it is just white noise, sorry.

Alexander

[Arthur Pemberton]

Lloyd Hayes wrote:
> Well, I got fed up with the bugs in Fedora Core 4 and put Fedora Core 3
> back on my computer.
>
> FC3 came out a year ago. One of the first things that I noticed was that
> some of newer programs did not work well with the Selinx policies.
> Updating the software took sever hours. (All night)
>
> This morning, after all of the downloading, I was almost required to
> install the new kernel first.
>
> After that, the bugs of FC4 showed up. They had not been there from the
> original kernel on the installation CD. They showed up after installing
> the new kernel. It appears that the bugs associated with FC4 are related
> to the new kernels being used. The same bugs show up in FC3 when the
> latest kernel is applied.

Hence why I try to tell people don't be too quick to blame probs solely 
on the distro. Unless you're a kernel hacker, file bug reports, and 
google for temporary work-arounds.

[thomas cameron]

On Sun, 2005-07-31 at 16:04 -0600, Lloyd Hayes wrote:
> Well, I got fed up with the bugs in Fedora Core 4 and put Fedora Core 3
> back on my computer.
>
> FC3 came out a year ago. One of the first things that I noticed was that
> some of newer programs did not work well with the Selinx policies.
> Updating the software took sever hours. (All night)
>
> This morning, after all of the downloading, I was almost required to
> install the new kernel first.
>
> After that, the bugs of FC4 showed up. They had not been there from the
> original kernel on the installation CD. They showed up after installing
> the new kernel. It appears that the bugs associated with FC4 are related
> to the new kernels being used. The same bugs show up in FC3 when the
> latest kernel is applied.

Was there a question there?

If you are seeing bugs, what are the Bugzilla numbers for your reports,
please?

Thanks,
Thomas

[taso]

Lloyd Hayes wrote:
> Well, I got fed up with the bugs in Fedora Core 4 and put Fedora Core 3
> back on my computer.

It is inevitable that Fedora users will occasionally get bitten by bugs. Their
purpose in life is to provide a diverse proving ground for new software so that
bugs are found and eliminated before said software is incorporated into Redhat
releases. The quid pro quo is that you get to use software which most of the time
works well enough to be useful.

If you like life right on the bleeding egde it is helpful to everyone if you
submit bug reports - the sooner they are documented and investigated the sooner
they get fixed.

[Thomas Taylor]

On Sunday 31 July 2005 15:04, Lloyd Hayes wrote:
> Well, I got fed up with the bugs in Fedora Core 4 and put Fedora Core 3
> back on my computer.
>
> FC3 came out a year ago. One of the first things that I noticed was that
> some of newer programs did not work well with the Selinx policies.
> Updating the software took sever hours. (All night)
>
> This morning, after all of the downloading, I was almost required to
> install the new kernel first.
>
> After that, the bugs of FC4 showed up. They had not been there from the
> original kernel on the installation CD. They showed up after installing
> the new kernel. It appears that the bugs associated with FC4 are related
> to the new kernels being used. The same bugs show up in FC3 when the
> latest kernel is applied.

Hmmm, there are bugs in my house, but just saying there are bugs won't make 
them go away.  In order to solve your problems with FC4, you need to provide 
some information (what kind of problem, which kernel, error messages, etc.).

Tom

[Timothy Murphy]

taso wrote:
> If you like life right on the bleeding egde it is helpful to everyone if
> you submit bug reports - the sooner they are documented and investigated
> the sooner they get fixed.

It's absurd to describe Fedora-4 as "bleeding edge".

I have quite a lot of sympathy with the OP
(though he could have described his problem more precisely).
Whoever is responsible for new Fedora kernels
tries to do far too much,
adding entirely unnecessary "improvements" to grub.conf .

It would be much better if he/she just installed the kernel
in the standard way with "make install",
and left it as an extra option for grub.

Actually, I don't understand why people find the Fedora kernels
so attractive.
They are rarely any improvement on the standard kernels
from http://www.kernel.org .

[John Wendel]

Timothy Murphy wrote:
> taso wrote:
> > If you like life right on the bleeding egde it is helpful to everyone if
> > you submit bug reports - the sooner they are documented and investigated
> > the sooner they get fixed.
>
> It's absurd to describe Fedora-4 as "bleeding edge".
>
> I have quite a lot of sympathy with the OP
> (though he could have described his problem more precisely).
> Whoever is responsible for new Fedora kernels
> tries to do far too much,
> adding entirely unnecessary "improvements" to grub.conf .
>
> It would be much better if he/she just installed the kernel
> in the standard way with "make install",
> and left it as an extra option for grub.
>
> Actually, I don't understand why people find the Fedora kernels
> so attractive.
> They are rarely any improvement on the standard kernels
> from http://www.kernel.org .

I've posted about this before, without any replies, so here I go again ...

I have been unable to get a kernel.org kernel to run correctly on FC4. 
The same kernel configuration works fine on FC3. The last kernel I tried 
was 2.6.12.5.

I'd appreciate any tips you might have for getting a stock kernel running.

Thanks,

John

[John Patrick Poet]

On Fri, 19 Aug 2005, John Wendel wrote:
> I've posted about this before, without any replies, so here I go again ...
>
> I have been unable to get a kernel.org kernel to run correctly on FC4.
> The same kernel configuration works fine on FC3. The last kernel I tried
> was 2.6.12.5.
>
> I'd appreciate any tips you might have for getting a stock kernel running.
>
> Thanks,
>
> John

When I upgraded from FC2 to FC4, but Kernel 2.6.12 to Kernel 2.6.12, I tried
using make oldconfig.  Like you, the resulting kernel would not boot.

When I went through from scratch using make xconfig, then the kernel worked
beautifully.

I do not know why...

John

On Mon, Aug 01, 2005 at 06:33:59PM +0100, Timothy Murphy wrote:
> taso wrote:
>
> It would be much better if he/she just installed the kernel
> in the standard way with "make install",
> and left it as an extra option for grub.

What is standard about that way of installing. It is rather primitive
in my mind to the rpm way.
> Actually, I don't understand why people find the Fedora kernels
> so attractive.
> They are rarely any improvement on the standard kernels
> from http://www.kernel.org .

=======================================================================
"And, you know, I mustn't preach to you, but surely it wouldn't be right for
you to take away people's pleasure of studying your attire, by just going
and making yourself like everybody else.  You feel that, don't you?"  said
he, earnestly.
    	-- William Morris, "Notes from Nowhere"
-------------------------------------------
Aaron Konstam
Computer Science
Trinity University
telephone: [phone]

[Timothy Murphy]

John Wendel wrote:
> > Actually, I don't understand why people find the Fedora kernels
> > so attractive.
> > They are rarely any improvement on the standard kernels
> > from http://www.kernel.org .
> I've posted about this before, without any replies, so here I go again ...
>
> I have been unable to get a kernel.org kernel to run correctly on FC4.
> The same kernel configuration works fine on FC3. The last kernel I tried
> was 2.6.12.5.
>
> I'd appreciate any tips you might have for getting a stock kernel running.

I don't set up as a great expert on kernel compilation
(or any computing subject, for that matter).
But I guess you are asking me ...

First of all, I gather your compilation went OK,
but the compiled kernel does not run.

If that is so, the problem probably lies in initrd.
My procedure, in brief, is
        make oldconfig xconfig
        make
        make modules_install
        make install

The last step runs mkinitrd (under Fedora),
installs the initrd*.img in /boot ,
and adds an appropriate entry to /etc/grub.conf .

I think sometimes mkinitrd does not choose the right modules.
In that case you could run linux under another kernel
(assuming you have one that runs),
add the required module(s) to /etc/modprobe.conf
and then run mkinitrd yourself.

Alternatively, some modules -
such as ext2/ext3 and scsi if you have SCSI disks - 
are better included in the kernel, IMHO,
though I think in theory it should always be possible
to include them in the initrd.

Final remark: I always save all my old .config files,
and load the most recent one when I run "make xconfig".
This has become very simple with (fairly) recent kernels,
with the window that comes up when you "make xconfig"
being much easier to use than it used to be.

I always say "make oldconfig xconfig",
and just answer Yes to all the questions I am asked,
and then load my previous config file.
I'm not really sure if there is any merit
in saying this rather than just "make xconfig".
I think there were a couple of kernels some time ago
when there were problems if one did not do this.

I'd be interested to know if there is in fact
any point in still saying "make oldconfig xconfig"
(assuming one is going to load an old config anyway),
or if it is just a fetish?

Re the title "Bugs in new kernels"
I've compiled most new kernels for some time,
and haven't encountered anything I would describe as a bug.

[John Wendel]

Timothy Murphy wrote:
> John Wendel wrote:
> > > Actually, I don't understand why people find the Fedora kernels
> > > so attractive.
> > > They are rarely any improvement on the standard kernels
> > > from http://www.kernel.org .
> > I've posted about this before, without any replies, so here I go again ...
> >
> > I have been unable to get a kernel.org kernel to run correctly on FC4.
> > The same kernel configuration works fine on FC3. The last kernel I tried
> > was 2.6.12.5.
> >
> > I'd appreciate any tips you might have for getting a stock kernel running.
>
> I don't set up as a great expert on kernel compilation
> (or any computing subject, for that matter).
> But I guess you are asking me ...
>
> First of all, I gather your compilation went OK,
> but the compiled kernel does not run.
>
> If that is so, the problem probably lies in initrd.
> My procedure, in brief, is
>         make oldconfig xconfig
>         make
>         make modules_install
>         make install
>
> The last step runs mkinitrd (under Fedora),
> installs the initrd*.img in /boot ,
> and adds an appropriate entry to /etc/grub.conf .
>
> I think sometimes mkinitrd does not choose the right modules.
> In that case you could run linux under another kernel
> (assuming you have one that runs),
> add the required module(s) to /etc/modprobe.conf
> and then run mkinitrd yourself.
>
> Alternatively, some modules -
> such as ext2/ext3 and scsi if you have SCSI disks - 
> are better included in the kernel, IMHO,
> though I think in theory it should always be possible
> to include them in the initrd.
>
> Final remark: I always save all my old .config files,
> and load the most recent one when I run "make xconfig".
> This has become very simple with (fairly) recent kernels,
> with the window that comes up when you "make xconfig"
> being much easier to use than it used to be.
>
> I always say "make oldconfig xconfig",
> and just answer Yes to all the questions I am asked,
> and then load my previous config file.
> I'm not really sure if there is any merit
> in saying this rather than just "make xconfig".
> I think there were a couple of kernels some time ago
> when there were problems if one did not do this.
>
> I'd be interested to know if there is in fact
> any point in still saying "make oldconfig xconfig"
> (assuming one is going to load an old config anyway),
> or if it is just a fetish?
>
> Re the title "Bugs in new kernels"
> I've compiled most new kernels for some time,
> and haven't encountered anything I would describe as a bug.

Thanks for the good advice.

The last kernel I built (2.6.12.5) will boot, but the network won't 
start. I didn't think to check the ram disk config for the proper 
network modules. I'll do that as soon as I can get to the machine.

Regards,

John

[Timothy Murphy]

[e-mail] wrote:
> > It would be much better if he/she just installed the kernel
> > in the standard way with "make install",
> > and left it as an extra option for grub.
> What is standard about that way of installing. It is rather primitive
> in my mind to the rpm way.

You misunderstood what I was saying.
Just before the bit you quoted above I said
"Whoever is responsible for new Fedora kernels
tries to do far too much,
adding entirely unnecessary "improvements" to grub.conf ."

My argument was that the "official" kernels
tinker too much with grub.conf ,
not that rpm was or was not a good way of installing a kernel.

Having said that, I run compiled kernels myself
because I use non-standard modules,
eg my USB PCI dongle did not work with the standard hermes module
last time I tried.

But I certainly wouldn't argue with anyone
who wants to run an "official" distribution kernel,
if it runs OK for him/her.

[Rahul Sundaram]

Hi
> Actually, I don't understand why people find the Fedora kernels
> so attractive.
> They are rarely any improvement on the standard kernels
> from http://www.kernel.org .

Well thats a deliberate design decision. Fedora packages as far as 
possible does not deviate from upstream ones. So when you come across a 
bug you can almost be sure that its a upstream one. Further every 
additional patch results in added work, so its important that developers 
evaluate the cost vs advantage of having it. However the kernel does 
have a few improvements like exec shield including several bug fixes 
parts of which has already been merged upstream.

regards
Rahul