Hi Aaron, Le 12/10/2022 à 04:48, Aaron M. Ucko a écrit :
Pierre Gruet <pgt@debian.org> writes:Indeed, this warning should be overridden, usually I do so and add a pointer to the discussion in https://lists.debian.org/debian-java/2018/06/msg00021.html in the comment in the .lintian-overrides file.Got it, thanks!Do you have an idea of the exact way the .so is loaded by the Java code? The methods in LibManager.java are a bit involved and I am unsure to see which behaviour is expected exactly. If you don't, I shall investigate further in order to help, no worry.No, sorry. That said, this issue is neither new nor urgent; I just noticed it and wanted to take the opportunity to address it if it would be reasonably easy to do so.
I have just pushed some changes. I adopted the -- I believe -- least invasive solution by creating a new package libngs-jni which depends on the shared lib package (not the -dev one), only ships a symlink to the shared lib with full version number, and on which the -java package depends. By the way I also made it Architecture: all and ensured it was binNMU-able.
No patch of the upstream code is required.As such, the new -jni package is not Multi-Arch: same as it ships the shared lib symlink in /usr/lib/jni. But if you think it should be, then we could install the symlink in /usr/lib/<triplet>/jni. This is less canonical regarding the Java policy but technically that should be OK.
Also, I have added a Lintian override for the embedded JS, as we discussed. Feel free to ping me if some issues show up later, but I expect none. Cheers, -- Pierre
Attachment:
OpenPGP_signature
Description: OpenPGP digital signature