FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > ArchLinux > ArchLinux General Discussion

 
 
LinkBack Thread Tools
 
Old 06-18-2012, 05:48 PM
Chris Lumens
 
Default newui observations on bare metal

> * Install Destination
> - Using a USB stick for install, it shows up in the drive list and
> is selected. clicking on it clears the 'select a disk' warning, but
> doesn't toggle its state. If you leave the spoke and return then it
> toggles when clicked as expected.

I would assume we are simply not taking protected devices into account
in the UI.

> - The 'review and custom' checkbox in the install options dialog gets
> set back on when you unselect it and then hit cancel/go-back. It
> stays clear if you continue and then revisit the spoke.

I think this is by design, but dlehman would have to confirm. I think
the idea is that once you've done the review, you won't want to again.
I've seen this before too and did think it was a little weird, so
perhaps it's worth revisiting.

> - The install options checkbox feels out of place to me. Instead of a
> single 'continue' button in the spoke we could instead have 2 buttons
> for review/custom or proceed.

I don't know the design rationale here but I would guess the idea is
that having more buttons is more confusing. Personally, I don't find
the checkbox to be out of place.

> - When in this spoke I am unsure what continue is going to do or where
> it will take me.

Maybe just a rewording of the button would clear it up?

> - There seems to be a resize happening when you go to this spoke, I can
> briefly see the upper right text left justified and then it jumps into
> place. I also see the disk selection section change size very quickly
> when entering the spoke.

I've seen this elsewhere, but have not looked into it yet. There are a
couple places where there's some weird resizing behavior that I'd really
like to take care of.

> * Custom Partitioning
> - On the second visit to the custom partition spoke I get a TB
> complaining about filling in the ProductName

I just reproduced this. I believe one of dlehman's patches that he
mailed today fixes it. I was just being sloppy here.

> - If I click on the create default layout I get a TB -- I think this
> is expected. It comes from kickstart execute() partitioning.py and
> throws a NoDiskError

Can you post the details of this somewhere? What was your existing disk
setup?

> - I don't like the progressive disclosure used in custom where I have
> to click on at least 3 things to see the details of a partition
> (Unknown->partition->custom...). I'd rather see all the details from
> the start.

Would having the expander expanded by default make things better? I can
see that being too busy-looking for the average user, though. Also, it
kind of makes having an expander pointless to begin with.

> * Installation Source
> - You can't select this until it finishes downloading the default
> metadata from the closest mirror. It takes enough time to download
> that you notice it. I think you should be able to enter the spoke
> and change the selection while the download is running -- eg. if
> you want to use a local repo instead.

I don't know that we can do this, though. The download isn't really
interruptable since it's running in yum code and we don't have any way
to kill it.

> - The closest mirror selection looks empty with the blank text
> field, maybe pre-fill it with whatever mirror was picked?

That might imply that you can change the context of the entry box, even
though it is shaded out. It does look weird, though.

> * Software Selection
> - After using the Install Destination with its back and continue
> buttons this spoke is a bit confusing with just a back button.

This is a quirk of Install Destination, not everything else. So far
it's the only spoke that can potentially take you off into a linear
sequence. I don't think we can do anything about it, though.

> * Install
> - This is still pretty rough, which is expected.

More should be functional. If you actually have ransom notes, they'll
get displayed on the bottom where there's currently a broken image.
We still have a lot of post-install configuration to do.

> - disk needs to be wiped and labeled for it to work right now

This is my guess as to why clicking on the autopart button failed for
you.

> - UEFI isn't working

Shocked, I say!

> * Misc
> - Anaconda is using 99% of CPU

You've taken care of this. Thanks.

> - no upgrade checks yet, no prompt to unlock LUKS (expected)

Everything upgrade-related is moving to preupgrade (or some other
non-anaconda tool) so there shouldn't be any upgrade checks to do.

LUKS, busted RAID, unlabelled disks are all known to not yet work. This
could be a "fun" task for someone to work on.

- Chris

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
 
Old 06-18-2012, 07:04 PM
David Lehman
 
Default newui observations on bare metal

On Mon, 2012-06-18 at 13:48 -0400, Chris Lumens wrote:
> > * Install Destination
> > - Using a USB stick for install, it shows up in the drive list and
> > is selected. clicking on it clears the 'select a disk' warning, but
> > doesn't toggle its state. If you leave the spoke and return then it
> > toggles when clicked as expected.
>
> I would assume we are simply not taking protected devices into account
> in the UI.
>
> > - The 'review and custom' checkbox in the install options dialog gets
> > set back on when you unselect it and then hit cancel/go-back. It
> > stays clear if you continue and then revisit the spoke.
>
> I think this is by design, but dlehman would have to confirm. I think
> the idea is that once you've done the review, you won't want to again.
> I've seen this before too and did think it was a little weird, so
> perhaps it's worth revisiting.

This is a result of the destination spoke being odd. Clicking "Back"
doesn't trigger a call to the apply method, so the settings are not made
permanent. IOW the back button on the storage spoke is currently a
Cancel button. This is due to everything in storage being opt-in. I'm
not sure what other approach we can take here without potentially
modifying someone's disk during OS install without them having
explicitly told us to do so.

> > - I don't like the progressive disclosure used in custom where I have
> > to click on at least 3 things to see the details of a partition
> > (Unknown->partition->custom...). I'd rather see all the details from
> > the start.

I would certainly like a device name and possibly type for each device
to avoid lots of clicking to identify things that weren't in an fstab
(and things like vgs and pvs).

This brings us to another question: when the user removes the last lv
from a vg, should we go ahead and remove the vg? How about its pv(s)? If
we don't, they get dropped down into the "Unused" group, which is not
ideal. It starts to feel like whack-a-mole.

>
> Would having the expander expanded by default make things better? I can
> see that being too busy-looking for the average user, though. Also, it
> kind of makes having an expander pointless to begin with.

I would personally like to be able to completely remove a specific
installation with one click.

>
> > * Installation Source
> > - You can't select this until it finishes downloading the default
> > metadata from the closest mirror. It takes enough time to download
> > that you notice it. I think you should be able to enter the spoke
> > and change the selection while the download is running -- eg. if
> > you want to use a local repo instead.
>
> I don't know that we can do this, though. The download isn't really
> interruptable since it's running in yum code and we don't have any way
> to kill it.

Either we don't do it automatically, forcing users to go in and choose
closest mirror, or we do it automatically (for those who do not provide
a repo) and potentially force some waiting while yum grabs metadata.
People will be complaining whichever way we go here.

Do we punish the procrastinator ("I want a non-default repo, but I
didn't want to type it at boot time") or the clueless user ("I don't
know what the hell this even means -- can't you just set it up for
me?")?

Rawhide metadata download can take a mighty long time, though.


_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
 
Old 06-18-2012, 07:15 PM
Chris Lumens
 
Default newui observations on bare metal

> This is a result of the destination spoke being odd. Clicking "Back"
> doesn't trigger a call to the apply method, so the settings are not made
> permanent. IOW the back button on the storage spoke is currently a
> Cancel button. This is due to everything in storage being opt-in. I'm
> not sure what other approach we can take here without potentially
> modifying someone's disk during OS install without them having
> explicitly told us to do so.

The fact that this Back button works completely differently from all
others is obnoxious, and I don't know what we could do. The button is
in the .c file and is not exposed through to the Python layer, nor do I
want it to be.

> > > - I don't like the progressive disclosure used in custom where I have
> > > to click on at least 3 things to see the details of a partition
> > > (Unknown->partition->custom...). I'd rather see all the details from
> > > the start.

> This brings us to another question: when the user removes the last lv
> from a vg, should we go ahead and remove the vg? How about its pv(s)? If
> we don't, they get dropped down into the "Unused" group, which is not
> ideal. It starts to feel like whack-a-mole.

>From the TODO at the top of pyanaconda/ui/gui/spokes/custom.py:

# - Removing a device is not very smart. It needs to take into account LUKS, LVM, RAID,
# all that kind of stuff. If this is the last device in one of those containers, all
# the containers should be deleted too.

So yeah, I agree - we should go ahead and remove it.

> > Would having the expander expanded by default make things better? I can
> > see that being too busy-looking for the average user, though. Also, it
> > kind of makes having an expander pointless to begin with.
>
> I would personally like to be able to completely remove a specific
> installation with one click.

Code-wise, it should be pretty easy to make the '-' button work when a
Page in the Accordion has the focus. We could hook into that and do the
delete there. I've got the button insensitive right now unless you've
got something displayed on the RHS. Getting rid of that would be a
first step.

Experience-wise, I think if the confirm dialog were sufficiently
helpful, this would be a nice way for people to get through the UI
quickly.

> > I don't know that we can do this, though. The download isn't really
> > interruptable since it's running in yum code and we don't have any way
> > to kill it.
>
> Either we don't do it automatically, forcing users to go in and choose
> closest mirror, or we do it automatically (for those who do not provide
> a repo) and potentially force some waiting while yum grabs metadata.
> People will be complaining whichever way we go here.
>
> Do we punish the procrastinator ("I want a non-default repo, but I
> didn't want to type it at boot time") or the clueless user ("I don't
> know what the hell this even means -- can't you just set it up for
> me?")?
>
> Rawhide metadata download can take a mighty long time, though.

Given that you can provide your repo via command line, kickstart, or
using the DVD I think we have plenty of ways for people to avoid an
unnecessary download. The way we do it now seems nicest to me for the
largest number of users.

- Chris

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
 

Thread Tools




All times are GMT. The time now is 07:16 PM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org