F21 Workstation System Requirements - i686

[Chris Murphy]

On Sep 5, 2014, at 3:39 PM, Owen Taylor <otaylor at redhat.com> wrote:

> On Fri, 2014-09-05 at 14:25 -0600, Chris Murphy wrote:
>> 
>> Will/should there be separate or slightly more qualified requirements for i686? Either hardware 3D support, or an age bracket?
>> 
>> I ask because I dug out a 10 year old Dell laptop from storage and it meets the current requirements: 2GB RAM, 60GB drive, 1.7GHz CPU.
>> 
>> But with AMD RV250/M9 GL FireGL 9000/Radeon 9000, the journal reports:
>> gnome-session-is-accelerated: No hardware 3D support. 
> 
> This is because of a blacklist
> in /usr/share/gnome-session/hardware-compatibility - I don't remember at
> this point why pre-R300 Radeon is blacklisted... as far as I know things
> *should* work. Theories:
> 
> * old Radeons typically have very small amounts of on-board video
> memory - 64MB is practically speaking pretty much the minimum that is
> going to work well.
> * Even several years ago when we created that file, we didn't feel like
> we had enough testing on R[12]xx to have any confidence in the drivers.
> * Some bug encountered in the past that may or may not have been fixed.
> * AGP, ugh, don't want to deal.
> 
> You can try commenting out the relevant line in that file and see what
> happens.

OK thanks.

> 
>> While gdm comes up, gnome-shell crashes (this is TC6). 
> 
> This is definitely not expected; since hardware support is disabled, you
> will have llmvpipe rendering, just like is used, e.g., when running
> Fedora in a VM. (i686 Fedora in a VM seems to work just fine.)

I failed to mention that gnome-shell comes up fine from the USB stick, I can run anaconda and install TC5, but the crashes happen on reboot from a successful install. So that's odd, but obviously a bug.
> 
>> With my QA hat on, I'd say if it worked, I'd use it against some of the
>> test cases; the fact it doesn't and i686 is (probably?) going away soon
>> anyway makes me think "if it works, bonus; if it doesn't, give up while
>> you still can" sort of attitude.
> 
> Our current thinking is that this system is marginal, but within the
> system requirements and any crashes are bugs we should fix.

OK good. 

Is it practical and beneficial to draw some line in the sand around "marginal" or "not supported", even if i686 hardware has 2GB RAM, and at least 10GB free space on the drive? Or is it literally anything i686 ought to work? (Allowing that maybe wireless might not work, or it may be slow due to no hardware 3D acceleration and just plain being really old.)


Chris Murphy