[ale] Fedora grousing

[Michael B. Trausch]

On 10/23/2013 08:10 AM, Jim Kinney wrote:
> As a   l o n g   time red hat then fedora user, I thought the new 
> installer in fedora 18 was bad implementation of a decent but 
> unpolished idea. 

I can't comment on Fedora 18's installer, as I've never used it.  I 
installed Fedora 17 and then used fedup to update to 18, 19, and then 20 
alpha.

However, due to a problem in 20 alpha (actually, several problems, but 
the primary one was its failure to interact with my FreeIPA domain) I 
had to roll my system back to Fedora 19, so I used the installer.

I found more than one problem with it, and I was actually going to start 
a thread about it, but since you've already done so...

My system is an MSI 970A-G46, which has UEFI as well as a BIOS which can 
run in the context of the UEFI firmware.  The firmware implementation 
is, so far, just fine.  However, Fedora 17 would /*not*/ install in UEFI 
mode, so I had to install it using the BIOS emulation.  That was less 
than ideal, but whatever.

Fedora 19 /did/ install in UEFI mode, /but it required two separate 
installs!/  Well, actually, /three/.  Ish.

*Attempt the first*:  Found that booting in UEFI mode didn't matter to 
the installer if the disk had an existing MBR but no GPT. Could have 
been handled in the installer code, but for whatever reason, it was 
not.  Not terribly surprising, though; I can see the assumption that the 
user wishes to keep the familiar unless they're installing on a 
brand-new disk to be reasonable.  So, I used parted and created a new 
GPT table on the device.  The installer refused to see the changes, 
though, so a restart of the installer was required.

*Attempt the **/second/*: Again, booted in UEFI mode. Selected Cinnamon 
desktop (it's what I've been using the most lately, outside of my 
experimental GNUstep environment) and all add-on features.  After about 
an hour or so of churning (egads, I _*/really/*_ hate the inefficiencies 
of the RH method of package management), the installer *crashed*.  Why?  
"Failed to remove the existing bootloader."  Duh, there wasn't one to 
remove.  Okay, well, fine. So, I spent another hour trying to get the 
mostly-installed system functional.  At this point, I really want to 
have a talk to the person that thought it was a good idea to run all of 
the post-installation tasks after the bootloader installation. /The 
bootloader installation should be the last blooody thing done, so that 
if it fails, the system can at least be bootstrapped manually and 
repaired./  Instead, the system had all its binaries and no 
configuration, and the amount of configuration in the post-install 
process performed by Anaconda is non-trivial, so...

*Attempt the /third/:* Again, UEFI booted the install media.  Overwrote 
the installation with just a Cinnamon install, no add-ons.  Waited 
another hour (/what?! I asked the installer to install nearly 800 less 
packages, and it took the same amount of time?!  Next time, I'm just 
going to go ahead and install them all anyway, then./) for the system to 
churn about and /this/ time the bootloader installation worked, 
post-install tasks ran, and I was on my merry frickin' way.

I *did* check out the manual installer.  I reasoned that it was about as 
usable as lead for disk partitioning, so I simply used parted.  Luckily, 
that can be done in a VT alongside the installer, but as I mentioned 
above, a restart of the installer is required for it to detect that 
you've made any changes to the media; simply running partprobe or kpartx 
isn't enough.

Sigh.

     --- Mike

-- 

	Michael B. Trausch

President, *Naunet Corporation*
? (678) 287-0693 x130 or (855) NAUNET-1 x130
FAX: (678) 783-7843

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ale.org/pipermail/ale/attachments/20131023/149be38c/attachment.html>