Updates to PKGBUILD for your consideration #456
Conversation
…4600). Removed unneeded deps (only need for building docs pacakge). Added prepare function for patching (new in pacman 4.1). Switched the git pull to a static zip file.
|
The zip file download is the way to go IMO. I actually requested a more formal versioning in upstream; we'll see what they think about it: raspberrypi/linux#262 |
|
What is inetutils needed for as a makedepend? |
|
Interesting, good to know. Also, have you built the package with these changes? |
|
Building now... 3+ hours and counting even with distccd... Have any of your devs gotten the Custom-built Linaro GCC to work under Arch? I currently cannot get the binutils package to compile: https://aur.archlinux.org/packages/arm-linux-gnueabi-gcc/ |
|
No, we use ct-ng. |
|
Why would you use those?
|
|
OK... I too am using ct-ng, with your toolchains but even with a pretty quick x86_64 client (i7-3770K @ 4.50 GHz), allowing the pi as the master to drive the compilation is very time consuming :/ |
|
That's why there are no Pi's in our build farm ;) |
|
Ha! So how long does kernel compilation take for you guys? |
|
45 minutes or so. |
|
45 min?! What am I doing wrong? Please review the post I referenced. I can tell you that I am seeing build jobs in my client's distccd.log but by no means is it smoking fast... it's been going since 6:45 PM (3h 20m) and still not finished. Relevant section of pi's /etc/makepkg.conf |
|
|
OK... so you guys have more powerful ARM boards... I assumed you were using the RPi too :p I would recommend compiling to tmpfs not to physical media. I have tmpfs on my workstation exporting to /srv/nfs/scratch which is where I compile. On workstation: On RPi: |
|
tmpfs on the devices is wasting RAM that generally needs to be used in building, and with SSDs the iowait is zero 99% of the time. And I'm way ahead of you with using it on the machine running ct-ng. |
|
Generally, yes. My workstation is 24 G of RAM... I allocate 14 G of it to tmpfs for scratch space so I'm in no danger of running out of memory :p Anyway, when you said, "And I'm way ahead of you with using it on the machine running ct-ng." Does that mean your are compiling with my pull request to verify that it works? I suspect you will finish hours before I do :) |
|
I meant that I've been using tmpfs for ct-ng compiles from near the beginning.. years ago. I don't compile pull requests, that's what the submitter is supposed to do before submitting. |
|
...which quad-core armv7 board are you using and would you recommend it or another model? I see several from which to choose. |
|
Both the Nitrogen6X/Sabre Lite and Odroid-X2/U2 are excellent.. both have a few pros and cons over each other so they about even out. The Exynos chips are faster though, and having 2GB of RAM isn't matched anywhere else. |
|
Cool. The ODROID-X2 looks really tempting. |
|
Man, that took a long time but I can confirm that it builds successfully and reboots into a functional system. If you wish to have a copy of the packages: |
|
This pull request will need to be refactored for changes due to reversion, it cannot be automatically merged at this point. |
|
You can probably reconstruct this in 10 sec. |
...0).
Removed unneeded deps (only need for building docs pacakge).
Added prepare function for patching (new in pacman 4.1).
Switched the git pull to a static zip file.