feat: allow building for unsupported architecture

The hard check here has been prevented us (the RISC-V port) from utilizing pkgctl build for a while. As long as the list (DEVTOOLS_VALID_ARCHES) isn't easily extendable, I think it's a better idea to allow building with a warning when specifying --arch manually.

Also disables repository detection and --repo check for unsupported architectures.

src/lib/build/build.sh

@@ -124,6 +124,7 @@ pkgctl_build() {
 	local RELEASE=0
 	local DB_UPDATE=0
 	local INSTALL_TO_HOST=none
+	local UNSUPPORTED_ARCH=0
 
 	local REPO=
 	local PKGVER=
@@ -165,7 +166,8 @@ pkgctl_build() {
 					BUILD_ARCH=("${DEVTOOLS_VALID_ARCHES[0]}")
 				elif ! in_array "${2}" "${BUILD_ARCH[@]}"; then
 					if ! in_array "${2}" "${DEVTOOLS_VALID_ARCHES[@]}"; then
-						die 'invalid architecture: %s' "${2}"
+						warning 'unsupported architecture: %s' "${2}"
+						UNSUPPORTED_ARCH=1
 					fi
 					BUILD_ARCH+=("${2}")
 				fi
@@ -312,6 +314,9 @@ pkgctl_build() {
 
 	# Update pacman cache for auto-detection
 	if [[ -z ${REPO} ]]; then
+		if (( UNSUPPORTED_ARCH )); then
+			die "Unsupported architecture specified, cannot auto-detect repository"
+		fi
 		update_pacman_repo_cache multilib
 	# Check valid repos if not resolved dynamically
 	elif ! in_array "${REPO}" "${DEVTOOLS_VALID_REPOS[@]}"; then
@@ -343,7 +348,7 @@ pkgctl_build() {
 		fi
 
 		# fail if an existing package specifies --repo
-		if [[ -n "${repo}" ]] && [[ -n ${pkgrepo} ]]; then
+		if [[ -n "${repo}" ]] && [[ -n ${pkgrepo} ]] && (( ! UNSUPPORTED_ARCH )); then
 			# allow unstable to use --repo
 			if [[ ${pkgrepo} == *unstable ]]; then
 				repo=${pkgrepo}

[Levente Polyak]

I don't really know much about the current riscv setup, but maybe you can help me understand why you have no pacman repositories that can be fetched and the repo detected? Is this a cross-platform build?

[Felix Yan]

The obvious problem here is that we don't have [multilib] in the RISC-V port (should also be the case for any new architectures that don't really need to enable 32-bit support), so the following command to fetch and update [multilib] will fail.

[Levente Polyak]

I see, so as we need to fix this anyway, I'd need to think a bit how we want to approach this for different architectures. It would be more future proof variant instead of more special cases around code