Re: [OE-core][PATCH 2/3] rpm-sequoia: New recipe for version 1.6.0

["Böszörményi Zoltán" <zboszor@gmail.com>]

2025. 01. 29. 10:29 keltezéssel, Alexander Kanavin írta:
> On Wed, 29 Jan 2025 at 10:20, Böszörményi Zoltán <zboszor@gmail.com> wrote:
>>>> +LICENSE = "LGPL-2.0-or-later & Apache-2.0 & BSL-1.0 & MIT & Unicode-DFS-2016 & (Apache-2.0 | MIT) & (MIT | Apache-2.0 | Zlib) & (Unlicense | MIT)"
>>> This is very elaborate, how was it determined? How to ensure it remains correct?
>> It was taken from the Fedora specfile license line verbatim.
>> IIRC, they (Red Hat) have a large contingent of lawyers
> I guess you could add a recipe comment pointing to the original specfile online.

The src rpm build from koji can be linked, I haven't found a link for
a separate specfile.

>
>>> Why can't we use upstream's standard install procedure? Also,
>>> cargo.bbclass has an install function, why is it completely replaced?
>> Because it fails with:
>>
>> | DEBUG: Python function extend_recipe_sysroot finished
>> | DEBUG: Executing shell function do_install
>> | ERROR: Did not find anything to install
> Yes, cargo_do_install only copies executables, you need to opt into
> libraries by setting CARGO_INSTALL_LIBRARIES.

It installs the library at the wrong location, but do_install:append
can still fix it so it matches the pkgconfig  file.

> But before that, presumably sequioa upstream has a standard
> installation procedure that we can try to use instead of copying files
> by hand?

It doesn't. It's README.md used LD_LIBRARY_PATH for
the example rpm build using it.

> I'm particularly interested in how ${S}/target/release/rpm-sequoia.pc
> is supposed to be installed.

It doesn't get installed, just generated.

I guess that's why the rust-rpm-sequoia specfile copies both
the so and the pc files explicitly in the correct location.