On Fri, Sep 18, 2015 at 1:52 AM, Dirk Hoffmann <[log in to unmask]> wrote:
>
>> Reporting "someone else had a probem, but I don't know how it
>> happened" is one of the great problems of debugging. It's not clear,
>> for example, if your colleague was trying to install on top of
>> pre-installed operating system on the new disk, or if they're familiar
>> with what is, admittedly, the horrible installation GUI form upstream
>> with its new and confusing "spoke and wheel" model.
>
>
> I am sorry, but that general place complaint is not appropriate here. I can
> provide all information that is needed or requested. I just do not claim
> having done all the tedious and repetitive work my own. If that hero does
> not want to subscribe to the list and make himself known, please respect it.
> We do not seek help here, but want to indicate a problem to help others.

It's very difficult to verify or get details from your complaint. And
now you seem upset because I called the difficulty into question.
There's missing critical information, such as whether your colleague
made the correct selections in the disk partitioning interface for SL
7, which is admittedly a confusing interface, or what other changes
they may have made in "creating a dumb partition", or even what
fileysystem that "dumb partition" uses.

I'm not trying in any way to say your colleague is not having a real
problem, and it might indeed be an installer software problem.


> There was a system running (SL6.x, depending on which of the three test
> machines it happened). However, on the new disk there was no partition table
> (as explained). We did not want to overwrite anything of the precious old
> configuration. And the only change between "Hangs somewhere in installation

Ah-ha. See, that part was vital, and was missing.

> with dark screen and (moving) mouse pointer" and "works" was to "fdisk" just
> one dummy partition on that new disk.

And, you can't run "fdisk" on a new disk without setting a label on
the new disk, unless the disk already had a label pre-set. Many do:
was the label pre-set to something oddball? Was the "new disk" a brand
new disk, or a pre-formatted disk from some vendor? Does it work if
you "zero" the disk, or at least zero the first 2048 blocks, which is
one of my favorite tricks for clickly restoring a disk to a "pristine"
state?

> I cannot remember the spoke and wheel GUI upstream. Do you mean the new
> interface for partitioning, which is already in service on Fedora 22 for
> example? I know that, but I do not see your point or the relationship to our
> problems.

I mean the new anaconda installer for SL 7, which is relatively new.
The "spoke and wheel" confuses the ordered checklist of the older,
more straightforward menus. And the partitioning part of that
interface can be *very* confusing. It directly violates Eric Raymond's
old "Luxury of Ignorance" guidelines for open source interfaces,
specifically guideliines 1, 2, 4, 6, and the first, thired, and fourth
of the postscript guidelines Eric added later.

>> I'd strongly urge you and your colleage to walk through a VM
>> installation and make sure you both know or learn how to manage
>> partitioning with advanced options, for just this sort of case.
>
>
> Why? What exactly do you think we need to understand? There was no
> possibility (at least not visible enough) to partition the new disk during
> the installation procedure. We guessed the reason and managed to partition
> it on another system manually (gparted).

The reason you guessed sounds unlikely. I'm not saying it didn't
happen, but that it seems unlikely, so I wonder what else was going
on. Did the disk not show up *at all* in the interface until a dummy
partition was created? How was the dummy partition created? Was a
label needed and added? Was there already partition information, such
as spurious LVM or other partitons?

> As a conclusion, I understand there is no requirement to have a
> pre-partitioned disk for installation (as expected and reasonable). Thanks.

And this is definitely not the case, I've verified it in the last 24
hours with VM based installation and entirely new disk images. You've
come to a conclusion inconsistent with other people's experience and
unlikely in the extreme for a freeware rebuild of a commercial grade
operating system.

> Can anybody reproduce the problem? We did it three times on different
> systems. And yes, if time allows I will try to reproduce it on a VM
> installation, but I am afraid VM disks (on VirtualBox) are or "look to the
> installer like" pre-partitioned disks anyway.

They really don't. They have neither disk labels nor partitions. Feel
free to verify this yourself. Now if the new disks you're dealing with
have some other weird setup, I'm quite curious. And if you can spare
the cycles, I'd be very curious if you can "zero the disk", or lat
least the beginning of the disk, and try a re-installation. This
*should* be easy to do with the installation DVD or netinstall CD with
the "rescue" mode, or with any live DVD.

> Cheers
>                                                                         Dirk