It looks like only ant fills manifests like this, using System.getProperty("java.runtime.version").

Since Java 9, openSUSE puts architecture in the optional part of the version:

> bash ../configure \
>    --with-version-feature=%{featurever} \
>    --with-version-interim=%{interimver} \
>    --with-version-update=%{updatever} \
>    --with-version-patch=%{patchver} \
>    --with-version-date=%{datever} \
>    --with-version-build=%{buildver} \
>    --with-version-pre="" \
>    --with-version-opt="suse-%{release}-%{_arch}" \

So there are at least two possibilities:

1) Remove arch from the optional field of the Java's version so that ant won't put it in manifest.

For example: https://build.opensuse.org/package/rdiff/home:1Antoine1:branches:Java:Factory/java-11-openjdk?opackage=java-11-openjdk&oproject=Java%3AFactory&rev=4

2) Or patch ant so that it puts either a constant string or a refined string in manifest's "Created-by" instead of using Java's detailed version as is. This may also prevent the excessive rebuilds/publications due to "Created-by" change on Java minor update.

For example, using system property "java.version" (i.e. major version only), like the 'jar' command does: https://build.opensuse.org/package/rdiff/home:1Antoine1:branches:Java:packages/ant?opackage=ant&oproject=Java%3Apackages&rev=4

I tested that your patched ant helps getting rid of arch-diffs (apart from the normal mtime variations in the jar).
There is now

Created-By: 11 (Oracle Corporation)

Reducing the --with-version-opt variance would still be a good thing for other reasons, though.

(In reply to Bernhard Wiedemann from comment #2)
> I tested that your patched ant helps getting rid of arch-diffs (apart from
> the normal mtime variations in the jar).
> There is now
> 
> Created-By: 11 (Oracle Corporation)

Yes, I copied the behavior of the 'jar' command.

I can put a fixed string instead, in order to be completely version agnostic. I don't think this field is relevant anyway for jars packaged as rpms…

Just let me know what you prefer.

If it's fine for you I can submit the patch to Java:packages.

> Reducing the --with-version-opt variance would still be a good thing for
> other reasons, though.

I've no opinion about this - no idea who actually is interested in such a version string.

having a version 11 in there, is perfectly fine.

When comparing my local build of xmlgraphics-fop to the official Factory one, I also noticed another diff in the MANIFEST.MF:

-Build-Id: 20180405-120000-GMT (abuild [Linux 4.15.13-2-default amd64, ^M
- Java 10+46-suse-1.1-x8664, Target Java 1.6])^M
+Build-Id: 20180405-120000-GMT (abuild [Linux 4.18.5-1-default amd64, J^M
+ ava 10.0.2+13-suse-1.1-x8664, Target Java 1.6])^M

capturing the user, kernel version and again java version details.
I wonder if anyone ever needs that...

(In reply to Bernhard Wiedemann from comment #4)
> having a version 11 in there, is perfectly fine.

I've submitted the patch, it's in staging now.

> When comparing my local build of xmlgraphics-fop to the official Factory
> one, I also noticed another diff in the MANIFEST.MF:
> 
> -Build-Id: 20180405-120000-GMT (abuild [Linux 4.15.13-2-default amd64, ^M
> - Java 10+46-suse-1.1-x8664, Target Java 1.6])^M
> +Build-Id: 20180405-120000-GMT (abuild [Linux 4.18.5-1-default amd64, J^M
> + ava 10.0.2+13-suse-1.1-x8664, Target Java 1.6])^M
> 
> capturing the user, kernel version and again java version details.
> I wonder if anyone ever needs that...

https://build.opensuse.org/request/show/645175

All the patches mentioned have now reached Factory. Is there anything left to do here or can we close this bug report?

I checked 2 packages from the list and build-compare finds them similar enough.

Minor things like varying mtimes in .jar files and date in javadoc .html
can be addressed later / long-term.

This is an autogenerated message for OBS integration:
This bug (1110024) was mentioned in
https://build.opensuse.org/request/show/791181 Factory / xmlgraphics-fop