Options to update (old, slow) laptop

[[n00b@localhost]]

Since my current laptop has broken and is irreparable I am having to use my old one until my work can order me a new one.

My current laptop, old laptop and workstation all run Gentoo. My old laptop has not been updated since September 2009 and, while it is possible to update a Gentoo installation that old, it would take too long so I am planning on wiping it and reinstalling from scratch (the partitions are currently ext3 as well so this would give me the opportunity to upgrade them to ext4 and also use disk encryption since this is a mobile system).

My old laptop is a IBM Thinkpad T22 with a 32-bit 800MHz Pentium 3 and 512 MB RAM. My workstation is a 64-bit 3.2GHz Core i7 with 6GB RAM. I would like to use my workstation to compile a complete Gentoo installation for my laptop but am unsure how to go about it.

Current options are:

1) A VM.
This would allow me to create an image file on my workstation which I could dd onto my laptop hard drive. The drawbacks are that it would use the toolchain optimised for the P3 to compile the system for the P3 and not the native toolchain optimised for the i7. It would also incur the overhead of a hypervisor which seems unnecessary since the i7 can run anything the P3 can.

2) A 32-bit chroot.
This would allow me to compile a complete 32-bit system inside a directory on my workstation which I could then rsync/tbz (stage 4) to copy onto my laptop (or loopmount an image as above). The drawbacks are again that the i7 would be using a 32-bit toolchain optimised for the P3 to compile for the P3 instead of the native toolchain optimised for the i7.

3) Crossdev.
This would allow me to create a 64-bit toolchain optimised for the i7 targeting 32-bit code for the P3. I am unsure how to integrate this with portage to create a complete 32-bit installation inside a directory though.

4) Prefix.
This would allow me to create a (partial?) Gentoo installation inside a directory on my workstation. I don't know if it is possible to create a 32-bit prefix on a 64-bit system however or even if a prefix system would boot when installed directly onto a hard drive. Also, having tried this before I am aware that not all packages in portage support being installed into a prefix.

DistCC is not an option as I don't think it is compatible with all packages required for a complete system and I don't want the laptop to do any compiling at all. Taking the laptop drive out and installing it in my workstation would work as well but as my workstation belongs to my work I don't think they'd be too keen on this option!

Are there any options I've missed? Has anyone successfully tried this before?

[dol-sen]

Option 1, would let you build and test the kernel and a complete system. Later it would let you build the binpkgs that can be emerged on the laptop to update it.

Option 2 would let you do all but the kernel. It would allow you to build binpkgs for later updates too.

Option 3, I don't know enough about.

Option 4, is a no, Prefix uses a different tree, paths, etc.
_________________
Brian
Porthole, the Portage GUI frontend irc@freenode: #gentoo-guis, #porthole, Blog
layman, gentoolkit, CoreBuilder, esearch...

[Veldrin]

personally I would go with the 32-bit chroot - takes the least effort.

the argument that the i7 has to run a p3-optimized toolchain (and implicitly is slower) does not really count as to compile most instructions are integer bases, and there only the i686 instruction set counts. the newly added instructions with p2/p3 are for multimedia applications, and less interesting for compiling.


I have to agree with dol-sen, that you cannot test the kernel, but you cannot do that either in a VM (virtualized hardware and such)


V.
_________________
read the portage output!
If my answer is too concise, ask for an explanation.

[[n00b@localhost]]

Thanks for the replies. I was leaning towards the 32-bit chroot anyway but was wondering if there was anything I missed. I had resigned myself to having to compile the kernel on the laptop (for other reasons) so one package out of ~1000 isn't too bad. I can also keep the chroot up to date if I want to reformat any partition on the laptop or perform binpkg installs. Also if I get an external HDD caddy later I can perform updates in the same manner chrooting onto the external HDD.

I disagree about the differences between the i7 and P3 toolchain. As an example there is an optimised strlen function in glibc which uses SSE4/AVX instructions on the i7 which are unavailable on the P3. I'm pretty sure gcc uses strlen but the number of times it does would be small enough to make the difference in performance negligible when compared to using a gcc/glibc optimised for the i7. Also due to increased core count, cache sizes, etc. the i7 will still be faster using the slower strlen.