[conspire] Elise's laptop

Rick Moen rick at linuxmafia.com
Sun Sep 10 17:44:47 PDT 2017


Quoting Daniel Gimpelevich (daniel at gimpelevich.san-francisco.ca.us):

> Funny that you put the exact processor model in your post, yet say that
> it is Bay Trail, and nobody but Dana thought to question this. For those
> following at home, he found that the N3050 is Braswell, not Bay Trail.

I could and should have included the fact that _everything_ I was saying
was a quick recap of sundry Web pages I found that might or might not 
be an exact match to Elise's and may or may not contain errors (Caveat
Emptor, Cave Canem, Mind the Gap...).


[deletiamo, deletiamas, deletiamat, deletiamus, deletiamtis, deletiamant]



> This model had Realtek wireless, which was recognized out of the box,
> but could hardly pick up any kind of signals until Dana found that the
> hardware was capable of two antennae, and the Linux driver was
> attempting to use _only_ the wrong one. He mentioned some kind of kernel
> module option that corrected the issue, allowing it to connect without
> the Edimax.

Chachma!



> I had not looked at that article until now, and it seems to be about a
> unit that was not identical (Broadcom wireless vs. Realtek).

Sometimes, these differences are because the OEM is using various
wireless thingies in the motherboard's mini-PCI slot or similar.
Sometimes, it's because the OEM is slipstream-changing the wireless 
component in the classic and long-despised Chipset Du Jour mode of
operation popularised back in dinosaur days by Dell ne PCs Limited.
Other times, it's actually a per-model difference within a product line.

Anyway, there's a good reason you should never absolutely count on
chipset identity until you can have lspci & kin report it directly from
the hardware.


> In particular, GalliumOS is tailored specifically for hardware that
> shipped with ChromeOS, with a number of hardware tweaks to that effect.
> Through experimentation (below) we saw that the merit of an XFCE-based
> system lighter than Xubuntu.

Good thinking.

While you guys were working this out, I was mostly dealing with the
fresh French bread and OEMing the home-made pizza.

[LXDE vs. other DEs for 'lightweight' aspect befitting low RAM and a
relatively low-puissance CPU:]

> To measure how light in resources a system would need to be, I started
> by booting a 64-bit Ubuntu MATE Edition 16.04.3 live image. It seemed to
> support the hardware well, and Elise said it was quite a bit faster than
> the stock Windows 10, but still a bit too slow, and allowed only one
> Firefox tab to be open in the 2GB RAM. Next, I pulled up a 32-bit
> Lubuntu 16.04.3 live image, but Dana and I failed to get the EFI to boot
> it in Legacy BIOS mode, required for the 32-bit image. Next up, Rick had
> downloaded the 64-bit Lubuntu 17.04 "alternate installer" image, which
> installed but had inoperable networking, which among other things,
> prevented the installation of SynClient GUI for the touchpad. I figured
> that we already saw the network operate in 16.04 but not 17.04, so it
> would make sense to install 16.04, and we opted for a step up from LXDE,
> which is XFCE, coming full circle. After installation completed, I had
> time only to install the packages zram-config and intel-microcode before
> handing the rest off to Dana while Rick and I went to help Elise with a
> car problem.

Can't argue with success.

> Elise had also brought with her at least five (!) other computers, of
> which we had time to examine in depth only two. She had installed Ubuntu
> 10.04 (32-bit) on an HP ProBook with 4GB RAM, a 250GB disk, and a Core
> i5 CPU of the Sandy Bridge generation, where it could not drive half the
> hardware in the machine. With Ubuntu's switch to GNOME 3 from Unity next
> month, Dana decided to download a 64-bit image of Ubuntu GNOME Edition
> 16.04 onto a USB stick, which installed uneventfully. The other machine
> had 1GB RAM (apparently expandable to 2GB in theory), an 80 GB disk with
> no bootable OS, and the AMD TL-50 Turion64 CPU with an nVidia
> northbridge and integrated GeForce 6150 graphics, where Elise reported
> previous graphics corruption. Dana booted that same GNOME Edition stick
> on it, and there was indeed intermittent texture corruption using
> Nouveau. Given the limited RAM, Rick downloaded a Debian 9 image, from
> which the intention was to install a base system and then manually add a
> light GUI. However, Debian installed XFCE without being told, which to
> our amazement worked well. I added the ZRam script from the Debian wiki,
> and the last thing we did was add the b43 firmware for its Broadcom
> wireless. An additional two laptops were left largely untouched with
> Windows 7, and a Raspberry Pi was never retrieved from her car. It
> should be noted that other people brought just as many computers
> yesterday, if not more all together, and non-computer issues were
> addressed too.

Thank you both for the astute and devoted work, and you, Daniel, for the
recap.

Elise reports that she enjoyed the visit, so she's either a good bluffer
or had an OK time, and intends to be back for more installation fun and 
perhaps some more good cooking.

Meanwhile, she's also twice over a hero in my eyes, a school teacher
and _also_ a CERT volunteer (the term meaning in this case Community 
Emergeny Response Team, not Computer Emergency Readiness Team).






More information about the conspire mailing list