Bug#1106317: Bug#1106313: streamripper: FTBFS in testing/arm64: configure: error: /bin/bash ./config.sub aarch64-linux-gnu failed
- To: Lucas Nussbaum <lucas@debian.org>
- Cc: Michael Ablassmeier <abi@grinser.de>, 1106313@bugs.debian.org, 1106314@bugs.debian.org, 1106315@bugs.debian.org, 1106317@bugs.debian.org, 1106318@bugs.debian.org, 1106319@bugs.debian.org, Sergio Durigan Junior <sergiodj@debian.org>, Alexandre Detiste <tchet@debian.org>
- Subject: Bug#1106317: Bug#1106313: streamripper: FTBFS in testing/arm64: configure: error: /bin/bash ./config.sub aarch64-linux-gnu failed
- From: Santiago Vila <sanvila@debian.org>
- Date: Mon, 26 May 2025 13:17:52 +0200
- Message-id: <[🔎] 6f7d43a4-d387-44db-92a4-fe31effc8bce@debian.org>
- Reply-to: Santiago Vila <sanvila@debian.org>, 1106317@bugs.debian.org
- In-reply-to: <[🔎] aDRIAlTCS0ygLpyp@grub.nussbaum.fr>
- References: <aC9zqLVOz5ShWLER@grub.nussbaum.fr> <aC9zqLVOz5ShWLER@grub.nussbaum.fr> <aDQceZGUeCsT3jJZ@ndnd.dead-link.org> <aC9zqLVOz5ShWLER@grub.nussbaum.fr> <83b7e616-cf34-4141-a90f-1dbb85f93a33@debian.org> <aDQqfrmmvBZjYo3q@ndnd.dead-link.org> <[🔎] aDQviRXUscEe7wD8@grub.nussbaum.fr> <[🔎] 62d53334-bd37-4181-b57e-a98b31cf1efc@debian.org> <[🔎] aDRIAlTCS0ygLpyp@grub.nussbaum.fr> <[🔎] aC9zkC6xoMmKZ4RL@grub.nussbaum.fr>
El 26/5/25 a las 12:52, Lucas Nussbaum escribió:
Looking at cdbs' changes history between those two releases, this sounds relevant:
https://salsa.debian.org/debian/cdbs/-/commit/5efba11b142b73ffbfa61fff09cdacc3d6db52b9
but I don't understand the change itself.
I "sort of" understand the change:
The line saying
ifeq (,$(shell grep -xi 'Package: autotools-dev' debian/control))
means "if the package being built is not autotools-dev".
Naturally, the autotools-dev does not need their own config.sub/config.guess
to be replaced by the master copies in the same autotools-dev package,
so that's why we have an ifeq there.
So, the ifeq was useful when autotools-dev was using cdbs itself.
Because autotools-dev is not using cdbs (I wonder if it ever did), the code may be simplified.
However, the right simplification would be to remove the ifeq and
do what it did unconditionally, not removing the entire block.
Same for gnulib. If gnulib used cdbs in the past, it needed the ifeq
to check that the package being built was not gnulib itself.
I see that the package is orphaned. I'm going to act conservatively
and revert the change completely. There will be plenty of time after
trixie for such kind of optimizations.
Thanks.
Reply to: