Timezone fix in validateInternalTimestampIsNotOutdated()

[landreev]

What this PR does / why we need it:

In the intended API workflow the client using the sourceLastUpdateTime parameter would be reusing the timestamps obtained from the output of other Dataverse APIs, like /versions and /files, where they are always reported in UTC, ISO 8601-formatted. (with the "Z", for ex.: 2026-04-22T14:30:00Z; regardless of where the server lives).
This format is specified as required in the guides.

However, the scheme appears to work only if the server happens to live in GMT and save the timestamps in the database that are already UTC.
That happens to be the case with Jenkins, beta, qa and docker instances (and that's how the bug stayed unnoticed for a year). But it becomes a problem on servers that use different time zones. Such as our "real" servers, demo.dataverse.org and dataverse.harvard.edu that live in US Eastern. (It was confirmed that file metadata could not be updated in the SPA on demo because of this)

Which issue(s) this PR closes:

• Closes Fix a bug in validateInternalTimestampIsNotOutdated() #12354

Special notes for your reviewer:

The simplest, minimalist fix would be 3 characters total: change yyyy-MM-dd'T'HH:mm:ss'Z' to yyyy-MM-dd'T'HH:mm:ssX. i.e., from "ignore the literal 'Z'" to "parse ISO 8601 time zone", which recognizes "Z" as "+0000".
But I added some comments and rewrote a few sentences that describe the parameter in the API guide.

Branch rebased to squash the commits.

Suggestions on how to test this:

Testing is straightforward, you just need a Dataverse instance operating in a non-GMT time zone. If you have a working Payara etc. setup installed locally on your computer, then you are all set. Unless you live in London, etc.

If you are testing with a Docker environment, you should be able to configure it to run in US Eastern by adding this to your docker-compose-dev.yml:

--- docker-compose-dev.yml
+++ docker-compose-dev-us-east.yml
@@ -9,6 +9,7 @@
     restart: on-failure
     user: payara
     environment:
+      TZ: America/New_York
       DATAVERSE_DB_HOST: postgres
       DATAVERSE_DB_PASSWORD: secret
       DATAVERSE_DB_USER: ${DATAVERSE_DB_USER}
@@ -113,6 +114,7 @@
     image: postgres:${POSTGRES_VERSION}
     restart: on-failure
     environment:
+      - TZ=America/New_York
       - POSTGRES_USER=${DATAVERSE_DB_USER}
       - POSTGRES_PASSWORD=secret
     ports:

Rebuild your Docker env. completely from scratch.

I recommend to make sure to reproduce the actual bug in 6.10.1 and current develop; as there were some conflicting reports on how easy that was.
It should be possible to test the bug, and the fix by simply making the 2 API calls involved directly. The best way however is probably to run one of the RestAssured tests that cover the functionality, for example

mvn test -Dtest=FilesIT#testUpdateWithEmptyFieldsAndVersionCheck

The failure would look something like this in the output:

            "directoryLabel": "data/subdir1",
            "lastUpdateTime": "2026-04-22T19:54:20Z",
            "fileAccessRequest": true
        }
    }
}
{
    "status": "ERROR",
    "message": "Internal version timestamp 2026-04-22T19:54:20Z is outdated"
}
[ERROR] Tests run: 1, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 13.06 s <<< FAILURE! -- in edu.harvard.iq.dataverse.api.FilesIT
[ERROR] edu.harvard.iq.dataverse.api.FilesIT.testUpdateWithEmptyFieldsAndVersionCheck -- Time elapsed: 10.40 s <<< FAILURE!

Note that the time stamp described in the API error as "outdated" appears strikingly identical to the actual lastUpdateTime in the output of the GET API right above.

Does this PR introduce a user interface change? If mockups are available, please link/include them here:

Is there a release notes update needed for this change?:

Additional documentation:

[coveralls]

coverage: 24.886% (-0.001%) from 24.887% — 12354-timezone-bug into develop

[landreev]

It's still working with the server living on GMT (it just passed Jenkins tests).
I'll clean it up and also the language in the API guide about these timestamps, then move it into ready-for-review.

[landreev]

The simplest fix would be just this:

--- a/src/main/java/edu/harvard/iq/dataverse/api/AbstractApiBean.java
+++ b/src/main/java/edu/harvard/iq/dataverse/api/AbstractApiBean.java
@@ -488,7 +488,7 @@ public abstract class AbstractApiBean {
     }
 
     protected void validateInternalTimestampIsNotOutdated(DvObject dvObject, String sourceLastUpdateTime) throws WrappedResponse {
-        Date date = sourceLastUpdateTime != null ? DateUtil.parseDate(sourceLastUpdateTime, "yyyy-MM-dd'T'HH:mm:ss'Z'") : null;
+        Date date = sourceLastUpdateTime != null ? DateUtil.parseDate(sourceLastUpdateTime, "yyyy-MM-dd'T'HH:mm:ssX") : null;
         if (date == null) {
             throw new WrappedResponse(
                     badRequest(BundleUtil.getStringFromBundle("jsonparser.error.parsing.date", Collections.singletonList(sourceLastUpdateTime)))

... no need really to mess with the incoming ISO 8601 string.

[github-actions]

📦 Pushed preview images as

ghcr.io/gdcc/dataverse:12354-timezone-bug

ghcr.io/gdcc/configbaker:12354-timezone-bug

🚢 See on GHCR. Use by referencing with full name as printed above, mind the registry name.

[ekraffmiller]

I tested this in my local environment by changing the timezone to eastern in docker compose, as described. I reproduced it in the unstable image (created a dataset, then attempted to edit, which failed). I switched to the PR image and Edit Dataset worked.