Completed first 32-bit migration to gcc-5.4.0-r3

eccerr0r

I finally completed my first migration to gcc-5.4.0-r3 as the system compiler. The system is a single core 1.6GHz HT Atom 32-bit with 2GB RAM. Took 4 days or so of compiling, debug and sleep, and a lot of the problems turned out to be distcc problems.

4 days, almost about as long as it'd take to emerge -e world without dealing with distcc. I don't know if it would have ended up needing as much handholding as the migration upgrade path took.

Have most people already done the migration? Perhaps having the mixed 32/64 bit environment is what's killing me.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

NeddySeagoon · Posted: Sat Oct 07, 2017 6:45 pm Post subject:

eccerr0r,

I did my Acer One Atom N270 in July and switched it to my best guess at the /17.0/ profile just recently.
I didn't have any trouble with distcc either.

Thats my last 32 bit system, unless you count Raspberry Pis.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

eccerr0r · Posted: Sat Oct 07, 2017 7:02 pm Post subject:

I would have expected things to go fairly cleanly but no... things had to blow up on me.

I still don't get why my distcc is so flaky and keep getting invalid object files from my cross-compile enabled 64-bit machines (as they're the newer Core machines, they are by far the faster machines). Fortunately I have one VM 32 bit setup on one of the 64 bit machines using KVM so that was the only 32 bit helper I could count on.

I wonder if I need to resort to a 32-bit chroot or container running distcc, but those too will need to be updated, so not much advantage there.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

NeddySeagoon · Posted: Sat Oct 07, 2017 7:24 pm Post subject:

eccerr0r,

My helper is a 64 bit clean (no-multilib) amd64 system with a i686-pc-linux-gnu cross toolchain.
I don't pass -m32 to the native toolchain.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

eccerr0r · Posted: Sat Oct 07, 2017 8:55 pm Post subject:

Perhaps that may be the reason, my 64-bit boxes are multilib but they should have 64-bit compiled cross compile toolkit for 32-bit, so theoretically, yes I should have two ways to generate 32-bit binaries. And I would hope the cross compilers can't generate 64-bit binaries...

But distcc shouldn't be passing -m32 as it's coming from a 32 bit machine. Also, would -m32 binaries should be compatible with 32-bit machines?

Perhaps forcing extra -m32 -arch=i686 would coax it to do the right thing, even if the wrong compiler was inexplicably selected?

Back the the original topic, have other people found the 5.4.0 transition to be fairly clean regardless of architecture, or riddled with (distcc-rootcaused) problems?
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

NeddySeagoon · Posted: Sat Oct 07, 2017 9:12 pm Post subject:

eccerr0r,

Its been clean here for i686, amd64 and arm64, with the amd64 helping out both the i686 and arm64, using cross distcc.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.