On 27 February 2026 4:32:55 pm IST, "Adam D. Barratt"
<[email protected]> wrote:
>On Fri, 2026-02-27 at 14:15 +1300, Andrew McMillan wrote:
>> You are correct that I am running on amd64 and looking at a package
>> which was built on arm64.
>[...]
>> My point is that I don't believe anyone is building and uploading NEW
>> packages on i386 but increasingly people are doing so on arm64, and
>> so the search that lintian is doing through the list of _arch.changes
>> files it prints should really include _arm64.changes (and
>> realistically it could probably drop _i386.changes, but that's not
>> really necessary).
>
>As Nilesh alluded to in a previous response, Lintian isn't using a
>hard-coded list here that includes i386 and could have arm64 added to
>it.
>
>Rather, it's using the set of architectures that your system knows
>about. The only hard-coded entries are "multi", "source", and "all".
>Anything else is determined at run-time, specifically:
>
>- $DEB_BUILD_ARCH, if set
> - the result of "dpkg --print-architecture" otherwise
>- $DEB_HOST_ARCH, if set
>- the result of "dpkg --print-foreign-architectures"
>
>The relevant section of the code is
>https://salsa.debian.org/lintian/lintian/-/blob/master/bin/lintian?ref_type=heads#L880-909
Right. If you do not provide the path to changes file explicitly, lintian tries
to guess the name.
I think checking for foreign architectures and build arch is a sensible way to
guess changes file name. The list of archs is nor hardcoded anywhere.
If we were to add in such a list of archs to check, this ends up opening a can
of worms - what do we do for multiple changes file? Which ones do we prioritize
and so on. I am inclined to mark this as a wontfix.
If anyone of you have suggestions, please let me know.
Best,
Nilesh
