[Bug 154763] Grub is inadequate without Lilo for backup

Guy Fraser guy at incentre.net
Wed Apr 20 19:45:03 UTC 2005


On Wed, 2005-20-04 at 13:31 -0400, Paul Iadonisi wrote:
> On Wed, 2005-04-20 at 10:52 -0600, Guy Fraser wrote:
> 
> [snip]
> 
>   Because I note a hint of willingness here, I'll respond again.
> 
> > 1) If I spend the time to test GRUB on FC4t2, and discover the 
> > same problems still exist, that I have seen before and can not 
> > resolve the issue using the wonderful info documentation, are 
> > you saying you would then consider LILO to be needed.
> 
>   That is not the nature of testing.  Though if it were up to me, I
> personally would not rule out re-including lilo, you're leaving out a
> few important steps, here.
>   If you discover the same problems still exist and you can't resolve
> them with the documentation you need to:
> 
> a) Define what 'same problems' you are talking about.  This may mean
>    chipset information, mobo info, drive types (IDE, PATA, SATA, SCSI),
>    drive arrangement (primary master, secondary slave, scsi id 0, etc.),
>    and other pertinent information (software raid info with config files).
>    You may not need to post this much detail to the list, but save it for
>    a possible, future bugzilla entry.
> b) See if someone on the list can help.  If the problem is solved with
>    someone's help from the list, then it's probably a documentation bug and
>    should be filed as such.
> c) Monitor any bugs you file (the default in Red Hat's bugzilla is to send
>    you mail when it's updated) in particular for requests for more information
>    or a request to check out a particular update on your specific setup.  An
>    all important goal here is assist developers in reproducing the problem
>    on their own hardware if at all possible.

You do realize that while doing all these tests I will not 
be able to use my machine for anything else.

> 
> And keep this feedback loop going, minus the 'grub sucks, lilo rules'
> rants.  We know where you stand on that...no need to muddy up the
> bugzilla database with it.
> 
> The point is, if you're going to help test grub, at least make an
> attempt to help developers fix it.  If you instantly want Red Hat to
> abandon the idea of eliminating lilo as soon as you find a reproducible
> bug in grub, then that's not much help.  Well, you can 'want' it, but
> don't give up on helping to fix grub in the meantime.
I do not suggest abandoning GRUB, it works well for many people.

> 
> > 2) One of the problems with the documentation may be that it 
> > requires pinfo. If the documentation was put in HTML format 
> > it would be *MUCH* easier to use. I have heard many people 
> > who find pinfo very difficult to use, and prefer to avoid 
> > info documentation all together. I don't care for it myself 
> > but can use it, when necessary.
> 
> +1
> 
> Don't even get me going on info files.  It's just...twisted.
I am not sure I understand that comment. Were you suggesting 
that you don't care for info documentation either?

> 
> > 3) What can you do about getting LILO reinstated if it can be 
> > demonstrated that it will work in situations where GRUB will 
> > not.
> 
>   See above.  I will only see the justification for that if I start
> seeing specific bug numbers with specific controller model numbers and
> specific drive arrangements and a clear description of the technical
> problem.  And even then, only if the problem can't be fixed in time
> before, say the next test release?  Remember, though, you don't need to
> wait until then to test a fix, provided an updated boot.iso and
> *stage*img files are put out with the new grub embeded.  Or maybe only
> the updated rpm in the rawhide is necessary -- I'm not sure how anaconda
> works regarding that.  Either way, test, test, test ... and include any
> updates that show up in rawhide in that testing.
I would rather not install anything from Rawhide, I have been 
badly bitten by it in the past. I am willing to help, but I
am not willing to loose all the data I have to do it. I am 
also not comfortable having to be without a bootable machine 
for more than a week. RH should have the resources to provide 
the developers with a ATA133 and SATA controller card along 
with some drives to test with. Not all QA should be the burden 
of the Fedora test community.

I may even consider donating the ATA133 and SATA controllers, 
since I will need at least one new SATA drive to do these 
tests. Once the tests are over I will want to use all the 
drives, and will need another SATA controller because I will 
then have 5 SATA drives.

All along I have not been attempting to get rid of GRUB, I 
have just wanted to be able to reliably use my machine, 
without having to re-install whenever I add or remove a 
non-boot drive.

> 
> 
> -- 
> -Paul Iadonisi
>  Senior System Administrator
>  Red Hat Certified Engineer / Local Linux Lobbyist
>  Ever see a penguin fly?  --  Try Linux.
>  GPL all the way: Sell services, don't lease secrets





More information about the fedora-test-list mailing list