Re: SOLVED: Re: Newbie kernel building questions

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, 2003-08-04 at 12:17, Andrew Robinson wrote:
> Jean Francois Martinez wrote:
> 
> >On Sun, 2003-08-03 at 22:01, Andrew Robinson wrote:
> >  
> >
> >>>Here's the procedure you should be following instead:
> >>>http://www.redhat.com/docs/manuals/linux/RHL-9-Manual/custom-guide/ch-custom-kernel.html
> >>>
> >>>      
> >>>
> >>That proceedure did indeed allow me to build my own kernel successfully. 
> >>I even added read capability for NTFS filesystems. Thanks for the tip!
> >>
> >>    
> >>
> >After checking that doc I would say it is entirely unadequate.  It is
> >unadequate because there are plenty of caveats who are not mentionned.
> >For instance in 2.4 kernels there are several parms who must be set
> >correctly or you won't have a working console (and in 2.6 it will be
> >much harder).
> >
> At the risk of sounding stupid (not a stretch), what's a "working console"?
> 

One where it displays something on the screen and where it reacts to
keyboard.

> >
> >I also hope you don't believe the BS about optimizing through
> >recompilation.  One of my boxes is a 64 megs and today it is
> >far too short on memory.  Just for the sake of it I recompiled the
> >kernel and, as expected, got no appreciable difference in speed.
> >On a 64 megs, recompiling bringed a little over 1% more memory.
> >And on a 512 megs...
> >
> I realize compiling my own kernel is not going to get much better 
> perofrmance, simply because the shipped kernels are fairly well 
> optimized. I'm guessing modules make a difference as well. However, I 
> also know building a kernel is a skill a "real man" should have. I had 
> never done it before, so I was pleased with myself when my machine 
> actually booted with the kernel I built. :)

"Real men don't compile kernels, they write them."  Linus.

Quote is fake but I couldn't resist. :-)

But one of the problems with "the cult of the kernel recompiling" is
that it makes the life difficult for people supporting apps on Linux
since they can't assume that even basic functionalities haven't been
left out by mistake: in a doc of a proprietary app ported to Linux I
found "make sure network support is included".  This is a bit extreme
but, for instance there is a seemingly inocuous network option who is
required for a working DHCP.   And the help text for the option doesn't
mention DHCP.  

> 
> >Finally if a feature you need is missing, and unless it is something
> >you know is experimental/not solid, you should also fill a bug report.
> >
> >
> >  
> >
> The feature I wanted was NTFS support. With my dual-boot box, it is 
> convenient to be able to read all the partitions. I think the problems 
> with NTFS support are fairly well documented, so I'm not surprised it 
> was not turned on with the delivered kernel. Didn't think that would be 
> worth a bug report.
> 

The idea is that it is better to have one person at RedHat doing the
work than one thousand in the world duplicating effort and their bosses
noticing and thinking that Linux has a high TCO.   If you can think of
a good reason for RedHat not including a feature (like too unstable for
prime time) then don't report, otherwise think in the one thousand
people and report.

> Please keep the comments coming. I want education!
> 
> Andrew Robinson
-- 
Jean Francois Martinez <jfm512@xxxxxxx>


-- 
Shrike-list mailing list
Shrike-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/shrike-list

[Index of Archives]     [Fedora Users]     [Centos Users]     [Kernel Development]     [Red Hat Install]     [Red Hat Watch]     [Red Hat Development]     [Red Hat Phoebe Beta]     [Yosemite Forum]     [Fedora Discussion]     [Gimp]     [Stuff]     [Yosemite News]

  Powered by Linux