davidc502 wrote:1. I recommend Ubuntu Distro running on a machine. Yes, a VM Guest running Ubuntu runs just fine to compile images
2. Install the following packages on installed Ubuntu - gcc, binutils, bzip2, flex, python, perl, make,
find, grep, diff, unzip, gawk, getopt, subversion, libz-dev and libc headers.
It's been a while since I've done step 2, but there may be a package or 2 in the list that will take a little doing to find. Take your time and you should be able to find it through a little research.
3. Get the lede code -- Create a folder and call it lede. Next, In command line, cd to that folder and run the following commands.
a. git clone https://git.lede-project.org/source.git
b. cd source
c. ./scripts/feeds update -all
d. ./scripts/feeds install -all
e. make menuconfig
If all of the prerequisites have been installed properly from step 2, the "make menuconfig" will invoke the Lede Configuration menu.
It will take a couple of minutes, but read how to use the arrow keys and "hotkeys".
1. Choose Target System = Marvell Armada 37x/38x/XP
2. Choose Target Profile = Choose the Image you want to create (Cobra) (Caiman) (Shelby) (Mamba).
Then save and that will bring you back to a command prompt.
Then run the command make V=s
From there it will start compiling minimum options, but will be good place to start to see if it will compile completely. After it's finished compiling, just check the /source/bin/targets directory for your compiled image.
Start out doing that and see if it works for you.
P.S. If you want to, in the LEDE Configuration menu scroll down to LuCi and choose it so you have a Gui to log into.
Just wanted to add, this is a great post.
I stood up a VM this evening on Ubuntu 16.04.1, and of course the package list was a little different. - most of the other packages come by default
I was able to just use the base build-essential and some *-dev packages (and of course git hehe)
I am currently compiling from your little guide :-)
Few gotchas, like:
- don't run as root :-) (bad habit of mine)
- ./scripts/feeds update -all ( is actually -a, not -all)
- ./scripts/feeds install -all ( is actually -a, not -all)
and likely a few others....
but I am getting there. My next questions will be about picking and choosing commit to build off of / Kernels / Drivers / and of course "Packages" etc.... but I can research on my own as well. Just wanted to say thanks!
I assume since this is all scripted and serial, we are favoring cpu speed per core and not the amount of cores..... At least that's what I am seeing, just watching the process jump from core to core shooting to %98. (but only ever over one core)
davidc502 wrote:@cybrnook2002
Actually, the new build will be 'building' tonight and into the morning, so I should be able to start trying it out some time tomorrow. In the meantime I fell back to r2221 which is rock solid.
Haven't played around with cake sqm yet. Maybe I'm wrong in my thinking, but didn't think it would really help out since I'm fortunate enough to have a fiber connection. It's supper low latency, and handles pretty much what I throw at it. Originally, I had a 1 gig connection, but downgraded to 100mbps after realizing it was a bit more than I needed
We are on a similar connection (bandwidth wise), latency may be another thing as I am still on archaic Comcast coax :-) but at 105/15 (provisioned at 125/25). However, of course I fail dslreports buffer bloat challenge, which twists my ocd..... That is the core of my venture into cake.
On a side note, I have noticed two Easter eggs in the WRT1900ACS builds:
1. the fan script for the ACv1 is still in cron (schedule). I would assume a build for the ACS would not need this. But I guess this all started with the ACv1, so it's just inherited in all the builds at this point?
2. The default interface for SQM is set to eth1. Which, I believe on the ACS should be set to eth0. So be careful if you do start to test QoS.
(Last edited by cybrnook2002 on 2 Dec 2016, 19:25)