Build combined

[jburel]

This PR builds openmicroscopy

• with and without ZarrReader (SNAPSHOT)
• with and without Bio-Formats (SNAPSHOT)

An error currently occurs with Bio-Formats (SNAPSHOT). The problem is not there with the release version
@dgault is looking into it

[dgault]

This failure looks to be caused by the fact that the ZarrReader is not being built whilst using a SNAPSHOT version and therefore it is unable to locate its underlying dependencies

[jburel]

The idea is to be able to build BF only and/or ZarrReader only and not create a dependency between the 2

[dgault]

If Ivy has a dependency for a SNAPSHOT version for ZarrReader then you are going have to build it. Using a deployed release version should be ok however.

[jburel]

@dgault the failure happens with the combination BF-SNAPSHOT+ ZarrReader 0.3.0
The combinations

• BF 6.10.0 + ZarrReader 0.3.0 works
• BF 6.10.0 + ZarrReader SNAPSHOT works

[dgault]

An ome-poi PR has been opened to bump the commons-logging version which should then match across the stack. This will require a release of the poi component and formats-gpl.

The other remaining discrepancies between BF and OMERO dependencies are as follows:
org.slf4j:slf4j-api:jar
BF: 1.7.6
OMERO: 1.7.30

We can bump slf4j in BF to match 1.7.30 (see ome/bioformats#3844)

org.apache.httpcomponents:httpclient:jar
BF: 4.5.9
OMERO: 4.5.6

commons-codec:commons-codec:jar
BF: 1.11
OMERO: 1.12

Both of these components in BF come from cdm-core:

+- ome:formats-gpl:jar:6.10.0:compile
|  +- edu.ucar:cdm-core:jar:5.3.3:compile
|  |  +- edu.ucar:httpservices:jar:5.3.3:compile
|  |  |  +- org.apache.httpcomponents:httpclient:jar:4.5.9:compile
|  |  |  |  \- commons-codec:commons-codec:jar:1.11:compile

In OMERO the ivy dependency resolution looks like:
httpclient

Organisation	Name	Revision	In Configurations	Asked Revision
com.google.http-client	google-http-client	1.20.0	compile, runtime	4.0.1
org.openmicroscopy	omero-blitz	5.5.12	default, compile, runtime, master	4.5.6
edu.ucar	httpservices	5.3.3	compile, runtime	4.5.9
org.apache.calcite.avatica	avatica-core	1.15.0	compile, runtime	4.5.6
org.apache.httpcomponents	httpmime	4.5.6	compile, runtime	4.5.6

commons-codec

Organisation	Name	Revision	In Configurations	Asked Revision
org.apache.httpcomponents	httpclient	4.5.6	compile, runtime	1.12
org.apache.solr	solr-core	3.1.0	compile, runtime	1.4
org.apache.calcite	calcite-core	1.20.0	compile, runtime	1.12

com.fasterxml.jackson.core:jackson-annotations:jar
com.fasterxml.jackson.core:jackson-core:jar
com.fasterxml.jackson.core:jackson-databind:jar
BF: 2.9.6
OMERO: 2.9.8

In BF these components are coming from minio:

+- org.openmicroscopy:ome-common:jar:6.0.9:compile
|  +- io.minio:minio:jar:5.0.2:compile
|  |  +- com.fasterxml.jackson.core:jackson-annotations:jar:2.9.6:compile
|  |  +- com.fasterxml.jackson.core:jackson-core:jar:2.9.6:compile
|  |  \- com.fasterxml.jackson.core:jackson-databind:jar:2.9.6:compile

In OMERO the ivy dependency resolution looks like:
jackson-core

Organisation	Name	Revision	In Configurations	Asked Revision
io.minio	minio	5.0.2	null, compile, runtime	2.9.6
com.fasterxml.jackson.dataformat	jackson-dataformat-yaml	2.9.8	compile, runtime	2.9.8
com.fasterxml.jackson.core	jackson-databind	2.9.8	compile, runtime	2.9.8
org.apache.calcite	calcite-core	1.20.0	compile, runtime	2.9.8
com.esri.geometry	esri-geometry-api	2.2.0	compile, runtime	2.9.8
org.apache.calcite.avatica	avatica-core	1.15.0	compile, runtime	2.9.8
com.fasterxml.jackson.core	jackson-databind	2.9.6	compile, runtime	2.9.6

jackson-databind

Organisation	Name	Revision	In Configurations	Asked Revision
io.minio	minio	5.0.2	null, compile, runtime	2.9.6
org.apache.calcite	calcite-core	1.20.0	compile, runtime	2.9.8
org.apache.calcite.avatica	avatica-core	1.15.0	compile, runtime	2.9.8

jackson-annotations

Organisation	Name	Revision	In Configurations	Asked Revision
io.minio	minio	5.0.2	null, compile, runtime	2.9.6
com.fasterxml.jackson.core	jackson-databind	2.9.8	compile, runtime	2.9.8
org.apache.calcite	calcite-core	1.20.0	compile, runtime	2.9.8
org.apache.calcite.avatica	avatica-core	1.15.0	compile, runtime	2.9.8
com.fasterxml.jackson.core	jackson-databind	2.9.6	compile, runtime	2.9.0

[sbesson]

As per https://github.com/FasterXML/jackson/wiki/Jackson-Release-2.9.7, I think upgrading jackson-databind to 2.9.8 at least would make sense

[sbesson]

Good to see that the combined build can now pass without requiring a chain of low-level components release. A few questions:

• the main donwside of declaring ome-common is that it's another version to track while it is defined transitively as a dependency of omero-blitz/omero-gateway. Is that something that would be done as part of every omero-blitz/omero-gateway bump as long as a new version is available?
• there are at least two locations where dependencies are declared ivy.xml and components/tools/OMEROJava/ivy.xml. I am confused and unclear on the role of each file especially as omero-blitz/omero-gateway are declared in both places. Should OMEZarrReader also be declared in components/tools/OMEROJava/ivy.xml and would that suffice to solve the issue?

[jburel]

An alternative I have considered is only to inject it as part of the GHA build since it is usually not required during a standard build since it comes via blitz/gateway.

[jburel]

@sbesson the last set of commits

• removes the change in ivy file
• modifies the ivy file in the github action.
  This allows us to test the matrix without having a new dependency to handle in etc/omero.properties

[sbesson]

Overall, 👍 for being able to test the cross-repository workflow via GitHub Actions build without any modification to the source files used for release

A few inline questions

[sbesson]

I remember it was used during the investigation but what is the rationale for including the ome-common dependency?

[jburel]

When building using snapshot, the version of ome-common does not get picked up for some reasons.