Re: poudriere cross-compile problem.

On 4/2/26 14:13, Craig Leres wrote:
> In the past I've used poudriere on an amd64 host (14.4) to build
> packages for my pfsense box (arm64.aarch64). In order to use the same
> git hash as netgate, I build my poudriere jails from a local source
> tree, e.g:
> 
>     zfs create -o mountpoint=/usr/src/16-CURRENT -o compress=lz4
> tank/16-CURRENT
>     cd /usr/src
>     git clone https://git.freebsd.org/src.git 16-CURRENT
>     cd 16-CURRENT
>     git checkout c215eef34550
>     poudriere jail -x -c -j pfsense2607 -a arm64.aarch64 -m src=/usr/
> src/16-CURRENT -b
> 
> This workflow was briefly broken, for example to build with 44f3e9f7f6c9
> you had to apply these changes:
> 
>  
> https://cgit.freebsd.org/src/patch/?
> id=8d5a11cd0137d3ad70f6b06e063ea91a776d510a
>  
> https://cgit.freebsd.org/src/patch/?
> id=38e5564ecef53064c984d086a040c01100023f40
> 
> It's my understanding that everything should be working again but I'm
> not finding that to be the case. When I try to build a package with this
> newly built jail I see:
> 
>     zinc 117 # poudriere bulk -c -j pfsense2607 -p pfsense ports-mgmt/pkg
>     [00:00:00] Cross-building ports for arm64.aarch64 on amd64 requires
> QEMU
>     [00:00:00] Creating the reference jail... done
>     [00:00:01] Warning: !!! Jail is newer than host. (Jail: 1600011,
> Host: 1404000) !!!

The following may be contributing . . .

Running a 1600011 poudriere(-devel) jail world on a 1404000 boot kernel
is not supported: the 1600011 jail may use new/changed 1600011 or
15????? kernel functionality/inerfacing that 1404000 kernels do not support.

Not that you would want to, but: What is supported is to run a 1404000
poudriere(-devel) jail world on a 1600011 boot kernel. 1600011 has means
of supporting the older kernels behavior/interfacing for older programs.

The boot kernel and its vintage is always involved. To run a 1600011
poudriere(-devel) jail world should normally mean booting 1600011 or
later kernel --or having a lot of knowledge about compatibility with
earlier 16000?? kernels for everything you need to have working.

The official port-package build servers boot 16000?? kernels at this
point, be that for 13.5, 14.3, 15.0, or main [so: 16.0] jail use. (The
boot world likely matches the boot kernel, though I've no direct
evidence of the boot world vintages.)

>     [00:00:01] Warning: This is not supported.
>     [00:00:01] Warning: Host kernel must be same or newer than jail.
>     [00:00:01] Warning: Expect build failures.
>     [00:00:02] Mounting system devices for pfsense2607-pfsense
>     [00:00:02] Warning: Using packages from previously failed, or
> uncommitted, build: /usr/local/poudriere/data/packages/pfsense2607-
> pfsense/.building
>     [00:00:02] Mounting ccache from: /var/cache/ccache
>     [00:00:02] Mounting ports from: /usr/local/poudriere/ports/pfsense
>     [00:00:02] Mounting packages from: /usr/local/poudriere/data/
> packages/pfsense2607-pfsense
>     [00:00:02] Mounting distfiles from: /usr/ports/distfiles
>     [00:00:02] Copying /var/db/ports from: /usr/local/etc/poudriere.d/
> pfsense-options
>     [00:00:02] Setting up native-xtools environment in jail... done
>     [00:00:02] Raising MAX_EXECUTION_TIME and NOHANG_TIME for QEMU from
> QEMU_ values
>     [00:00:02] Copying latest version of the emulator from: /usr/local/
> bin/qemu-aarch64-static
>     [00:00:02] Appending to make.conf: /usr/local/etc/poudriere.d/
> pfsense-make.conf
>     /etc/resolv.conf -> /usr/local/poudriere/data/.m/pfsense2607-
> pfsense/ref/etc/resolv.conf
>     [00:00:02] Starting jail pfsense2607-pfsense
>     [00:00:02] Error: Unable to execute id(1) in jail. Emulation or ABI
> wrong.

I'm unsure if the above specific message would be a possible result of
what I referenced earlier above.

Can you try booting with a 1600011 (or later) kernel, leaving the rest
the same in order to see if the behavior chnages?

>     [00:00:02] Cleaning up
>     [00:00:02] Unmounting file systems
>     Exiting with status 1
> 
> However the older jails still work. This makes me suspect native-xtools
> is still not quite right (or something else has changed).
> 
> Any hints?
> 
>         Craig
> 
> 


-- 
===
Mark Millard
marklmi at yahoo.com