distcc/crossdev question

[skiwarz]

Happy Thanksgiving, Americans! And happy Thursday to everyone else!

I have an old laptop with an atom processor. It's getting hard to emerge everything on it these days, so I'm trying to set up some sort of binhost on my desktop machine. Both have amd64 architectures, although my desktop is an intel i7, so their features vary a bit
All I really want to do is have my desktop PC emerge a separate "world" (or some other set) using a second make.conf, and export the packages to a binhost instead of installing.

The three options I know of (and why I don't think they work in this situation) are:
1. After updating my desktop, use quickpkg to make a binary I can then pull in from my laptop. However, I would then need to have ALL of my laptop's packages also installed on my desktop... which is not ideal.
2. Use distcc to simply speed things up. This COULD work, however I have some weird power settings and scripts on my PCs to suspend/wake at certain times, which might interfere with distcc. Plus, I've read that linking operations still need to occur on the local machine with this method, which could still potentially exceed the specs of my laptop.
3. Use crossdev. I read through the gentoo wiki on this, and it seems like the tool is meant more for compiling on different architectures, not simply for a machine with a different make.conf. Particularly, you are supposed to run "crossdev -t x86_64-pc-linux-gnu" which seems like it will duplicate my current toolchain needlessly (and possibly causing conflicts).

EDIT: After trying to set up crossdev anyways, it flat out refuses:

[skiwarz]

Kinda buggy, but here's what I came up with for a workaround, unless somebody can point me in a better direction:

create a "config-root" at /var/db/repos/laptop (it could have been anywhere, but this seemed logical)
create etc/portage in that config-root
copy all my portage config files over from my laptop into the new config-root
create a pkgdir folder on my NAS
modify the make.conf in my new config-root as such:

• change -march=native to something more usable
  
• change makeopts
  
• change PKGDIR to point to my new pkgdir

created a new set at /var/db/repos/laptop/etc/portage/sets/laptop containing the packages I wish to build on my desktop
now, whenever I want to compile a new version, I need to:
emerge -a1qv --buildpkgonly --config-root=/var/db/repos/laptop --nodeps @laptop

The only problem with this method is that it won't handle update detection (emerge -u), since it doesn't know which versions I've emerged previously (only what is installed on my desktop). So when I try to "emerge -u" on my laptop, it will try to pull packages from my binhost that aren't there (and then try to compile them itself), until I manually emerge them.

[pingtoo]

Just an idea, You can use an docker container simulate your laptop. Or an VM.

[NeddySeagoon]

skiwarz,

The trival case when the build CPU can execute all the instructions of the target CPU is to create a chroot that contains the install you want to build for.
You share distfiles and the repos, so you don't need two copies of them but you keep the binary packages separate.

Ask the Atom CPU what -march=native means and put that into make.conf or it will build for the wrong CPU.

[Hu]

The traditional answer here is to use a chroot, which is like a Docker container or VM, but is lighter weight. Regardless of which of those three options is chosen, the desktop will need to build and install (in the isolated area) separate copies of every package the laptop needs, even if the desktop already uses that package. This has the minor virtue that at least the package is not installed on the desktop's root filesystem, but the drawback that packages which are used in both systems will be built twice.

[skiwarz]

Thank you Neddy and Hu, you guys answered my question perfectly. I'll probably work on setting up a chroot soon. But in the meantime, I found a halfway decent workaround using some portage trickery:

Using the configuration I listed in my last post, I emerge the packages I need on my build system, adding the options "--buildpkgonly" and "--usepkg=y". Then I set a cron job to emerge the packages I need every night. If a particular version already exists in pkgdir, portage will pull it from pkgdir (--usepkg=y), unpack it and repack it, then upload back to pkgdir (--buildpkgonly). This way, I always have the newest versions available in pkgdir for my laptop's use, while not wasting processing time rebuilding every day.
It's a little gimmick-y, but it will do until I get around to setting up a proper chroot.

Thanks again guys,
-skiwarz

[jesnow]

Consider running a gentoo distribution on your laptop that has a binhost. I have been having good luck with calculate linux, which I install on old laptops and vm's that I use to run (guess what) distcc for my main machines. Calculate is pretty full featured, and leverages all your gentoo knowledge.

Jon.

[skiwarz]