Hi!
On Sun, 2025-03-02 at 02:50:35 +0100, Santiago Vila wrote:
> Package: dpkg-dev
> Version: 1.22.15
> I was going to ask ftpmaster to remove the "hello-traditional" package,
> but decided to still make a "last release" so that it's archived in
> snapshot.d.o.
>
> To my surprise the package failed to build on hurd-i386:
>
> [...]
> Use of uninitialized value $v{"DEB_HOST_MULTIARCH"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_ARCH_ABI"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_ARCH_BITS"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_ARCH_CPU"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_ARCH_ENDIAN"} in concatenation (.)
> or string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_ARCH_LIBC"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_ARCH_OS"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_GNU_CPU"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_GNU_SYSTEM"} in concatenation (.)
> or string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_GNU_TYPE"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> Use of uninitialized value $v{"DEB_TARGET_MULTIARCH"} in concatenation (.) or
> string at /usr/bin/dpkg-architecture line 354.
> dpkg-source --before-build .
> dpkg-buildpackage: info: host architecture hurd-i386
> invalid host_arch hurd-i386 at /usr/bin/dpkg-checkbuilddeps line 122.
> dpkg-buildpackage: error: build dependencies/conflicts unsatisfied; aborting
>
> Is this a bug in dpkg-dev?
That would seem a bit strange as I do not recall recent changes that
would affect his.
In any case, just to make sure, I was unable to reproduce this on my
hurd-i386 VM. I even tried on an amd64 chroot to remove the
build-essential and gcc packages (which is where the arch comes from),
but those fail gracefully too.
> ( The package now has autotools-dev as its only build-dependency,
> which is clearly strange but not technically a bug )
This looks like the chroot might have issues? Perhaps the prefixed
fixchroot command in front of dpkg-buildpackage is interacting badly
with it? (Which seems specific to the hurd-* ports).
I'm CCing the debian-hurd list.
Thanks,
Guillem
