Usage of import commands and powershell command prevents cross-compile

[gbraad]

In ./build.sh there is a reliance on:

wsl-vpnkit/build.sh

Line 6 in a074e1c

USERPROFILE="$(powershell.exe -c 'Write-Host -NoNewline $env:USERPROFILE')"

and

wsl-vpnkit/build.sh

Lines 17 to 20 in a074e1c

	# reinstall
	wsl.exe --unregister wsl-vpnkit || :
	wsl.exe --import wsl-vpnkit --version 2 "$USERPROFILE\\wsl-vpnkit" $DUMP
	rm $DUMP

to determine and run specific Windows settings and commands.

[gbraad]

Would it be possible to break this up, or perhaps adopt a Makefile that can be used on a automated pipeline? More than willing to help with this. WDYT?

[josesa-xx]

Actually since build.sh is meant to be run inside WSL, line 6 could just be replaced with:
USERPROFILE=$(wslvar USERPROFILE)

[gbraad]

The build.sh serves two purposes. one of them is the actual build that happens in the Docker container, the other is the import into the WSL environment. Several things can be improved:

• container runtime independent; as I run builds on Podman
• add a Makefile; use make to perform the build
• and the build.sh doing a rebuild invoking make and the unregister and import

And you are right, as invoking 'powershell' to get an environment variable is mpt as clean as wslvar.

[josesa-xx]

I've made some changes to my current pull request #169 to split build and import into different scripts and also allow build with podman as alternative.

[gbraad]

Thanks, I'll check it out

…

On Fri, Dec 2, 2022, 07:11 Jose Sa ***@***.***> wrote: I've made some changes to my current pull request #169 <#169> to split build and import into different scripts and also allow build with podman as alternative. — Reply to this email directly, view it on GitHub <#171 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAAOZTKR2TLD3ZVBH3NSOLWLEWDZANCNFSM6AAAAAASOGG2HE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[gbraad]

@josesa-xx for gvisor-tap-vsock we are adding a systemd service instead as it is more robust.
For our podman-machine implementation we are already using a systemd patch, so with the newer release of 22H2 this is fixed for future setups.

[sakai135]

I'll close this issue since #169 seems to resolve the issue.