Hi Hans,
On Wed, Apr 28, 2021 at 3:47 PM Hans de Goede <hdegoede(a)redhat.com> wrote:
Also I hope it is ok if I pick your brain a bit on a java
packaging issue which I've been having.
I maintain a couple of java leave packages (games) + some deps
which AFAIK are only used by these games.
One of these deps (dom4j) has been FTBFS since F34:
https://bugzilla.redhat.com/show_bug.cgi?id=1923601
I've been looking into this and the actual problem seems to
be with Java 9 now including what once was the org.relaxng.datatype
except they did not just bundle it, they also changed where it
sits in the namespace to com.sun.tools.rngdatatype <sigh>
Just doing a s/org.relaxng.datatype/com.sun.tools.rngdatatype/
got me a bit further, but it seems that msv, which is a dep of
dom4j needs to be rebuild first with the same search-replace
done on it and the FTBFS bug of msv is stuck because of one
of its deps getting orphaned+retired :
https://bugzilla.redhat.com/show_bug.cgi?id=1923446
So I think I can fix this by:
1. Unorphaning jvnet-parent, which is the missing msv dep
You can unorphan/unretire it, but removing dependency on jvnet-parent
is another choice. Probably better choice as jvnet-parent is no longer
developed or maintained by upstream.
2. Do a s/org.relaxng.datatype/com.sun.tools.rngdatatype/ in msv,
rebuild
3. Do a s/org.relaxng.datatype/com.sun.tools.rngdatatype/ in dom4j, rebuild
relaxngDatatype was retired in Fedora. A replacement is
jaxb-relaxng-datatype package (built from jaxb source package), but it
uses a different namespace. Therefore steps 2 and 3 seem correct and
necessary.
And then either do this only for rawhide, or push all 3 modified
packages to F34 in a single bodhi update.
Mikolaj, does this sounds like a reasonable plan to you; or
should I approach this differently?
Yes, that sounds good. I would also double check to see whether
msv/dom4j are really needed by your packages.
Also if yes this is a reasonable plan any advice on also
pushing the fixed packages to F34, or not ?
I am in favor of F34 update. Users of dom4j/msv that do not rely on
relaxngDatatype-related functionality should be unaffected by the
update. Users that do rely on it would get it fixed.
Regards,
Hans