HDDS-14976. xml_to_md.py does not handle generated XML files

[adoroszlai]

What changes were proposed in this pull request?

dev-support/ci/xml_to_md.py reads only ozone-default.xml. It misses other, generated XML files, e.g. hdds-common-default.xml.

• Handle all -default.xml config files, but skip network-topology-default.xml (which is a different kind of config). Also skip ozone-filesystem-hadoop shaded jar files, which contain contents from several other jars, no own config.
• Further tweak default values to handle special value:

  ozone/hadoop-ozone/httpfsgateway/src/main/resources/httpfs-default.xml

  Lines 111 to 117 in c0d50fe

  	<name>httpfs.services</name>
  	<value>
  	org.apache.ozone.lib.service.instrumentation.InstrumentationService,
  	org.apache.ozone.lib.service.scheduler.SchedulerService,
  	org.apache.ozone.lib.service.security.GroupsService,
  	org.apache.ozone.lib.service.hadoop.FileSystemAccessService
  	</value>

https://issues.apache.org/jira/browse/HDDS-14976

How was this patch tested?

Generated config doc:

python3 dev-support/ci/xml_to_md.py hadoop-ozone/dist/target after.md

verified missing properties (e.g. ozone.tracing.*) are present.

https://github.com/adoroszlai/ozone/actions/runs/23983488947/job/69952049151

[Russole]

Thanks @adoroszlai for the patch. Overall, LGTM. I just have a few minor comments.

[rich7420]

thanks for the patch @adoroszlai

[Russole]

I noticed that some property names still contain trailing newlines in the generated markdown (e.g. wrapped to the next line).

This seems unrelated to this change, but it might be worth normalizing name as well when parsing the XML, similar to how description is handled, for example:

name = prop.findtext('name')
name = ' '.join(name.split()).strip() if name else ''

This would help prevent formatting issues in the generated table. For example, some entries are currently rendered like:

| `hdds.container.ratis.ipc.random.port
` | false | `OZONE`, `DEBUG` | Allocates a random free port for ozone ratis port for the container. This is used only while running unit tests. |
| `hdds.container.ratis.leader.pending.bytes.limit
` | 1GB | `OZONE`, `RATIS`, `PERFORMANCE` | Limit on the total bytes of pending requests after which leader starts rejecting requests from client. |
| `hdds.container.ratis.log.appender.queue.byte-limit
` | 32MB | `OZONE`, `DEBUG`, `CONTAINER`, `RATIS` | Byte limit for ratis leader's log appender queue. |

[adoroszlai]

Thanks @Russole for spotting this.

[Russole]

Thanks for the change. It looks good, LGTM!

[sarvekshayr]

LGTM +1.

[sarvekshayr]

Thanks @adoroszlai for the patch, @Russole and @rich7420 for the review.

[adoroszlai]

Thanks @rich7420, @Russole, @sarvekshayr for the review.