Fails when downloads.openmicroscopy.org is down

[jkh1]

The package shouldn't depend on an active internet connection, at least not without properly informing users, and definitely should not rely on a single server which today failed and renders RBioFormats unusable and by extension any R script using it.
EDIT: I see that the java library used to be bundled so I think reverting that change would be the solution.

[aoles]

Did you try the latest version? I think the described problem has been addressed in c87bba1.

[jkh1]

I have 0.0.56. I can't install the latest as it is also trying to access downloads.openmicroscopy.org

[aoles]

I'm sorry for the inconvenience. v0.0.57 does not require internet connection once installed, however, it still needs to access the download server during installation in order to fetch the Bio-Formats java library. I believe the problems with accessing downloads.openmicroscopy.org are temporary, so please try updating the package again later.

As the java library is a rather heavy resource which keeps updating relatively frequently, I currently don't consider moving back to the former approach of bundling it inside the package.

[jkh1]

OK. Thanks for the prompt reply. I was caught off guard as I expected that once installed the package wouldn't need anything. It might be worth having the package try to pull the java library from different sources in case one fails, maybe it could be downloaded from the GitHub repo or from an openmicroscopy.org mirror if there's such a thing.

[aoles]

Thanks! I'm not aware of any mirror, but in case you find one I'm happy to implement the suggested solution.

[sbesson]

HI @aoles @jkh1, we are currently working to restore the binary downloads. However, several aspects are still outside our control and we will update https://forum.image.sc/t/ome-resources-down-due-to-uod-outage/43957 accordingly.

Looking briefly at RBioFormats, assuming bioformats_package.jar is the main requirement, I would suggest to downloading it from the OME Artifactory. At the moment, the OME Artifactory is also down but we have redirected the service to Glencoe's artifactory which should include the latest Bio-Formats artifacts. Additionally, https://maven.scijava.org/ is a third location which I think mirrors these artifacts.

[jkh1]

Thanks @sbesson I was just looking around for a mirror but wasn't finding one.

[aoles]

Thanks @sbesson for chiming in!

The linked resources seem to have the artifacts arranged in a different way than the original downloads.openmicroscopy.org repo. Using them would require additional logic other than just replacing the URL in order to automatically download the required library version. Therefore I don't really see this as a quick solution, maybe it's best to just wait until the primary download site is up again?

[sbesson]

We are definitely hoping to get the downloads up as soon as possible. In case you are considering switching immediately or in a near-future, the canonical URL for the versioned JAR and the SHA1 should be:

https://artifacts.openmicroscopy.org/artifactory/maven/${artifact_group}/${artifact_name}/${artifact_version}/${artifact_name}-${artifact_version}.jar
https://artifacts.openmicroscopy.org/artifactory/maven/${artifact_group}/${artifact_name}/${artifact_version}/${artifact_name}-${artifact_version}.jar.sha1

[jkh1]

I think it would make sense to have at least one other independent source in case this kind of outage happens again.

[aoles]

Thanks, I will look later tonight into enabling artifactory as a backup.

[leej3]

It's not clear to me if this and #24 are still an issue. It is possible to disable the RBioformat-jar's update behavior as described here. Looking at the shell scripts in bftools it looks like this flag propagates through to the java call as -Dbioformats_can_do_upgrade_check=false. If it is still a problem it might be worth appending -Dbioformats_can_do_upgrade_check=false to the java.parameters internally to see if that stops the upgrade check.