fix(push-publishing): prevent non-Host contentlets from being bundled as .host.xml

[gortiz-dotcms]

Summary

• HostBundler.writeFileToDisk() now skips any contentlet whose content type is not a Host (e.g. Widgets or regular content related to a Site via relationship field), preventing them from being incorrectly written with the .host.xml extension
• ContentHandler adds a null-safety guard around hostAPI.find() as a secondary line of defense — logs a warning and skips host-only update steps instead of NPE-ing
• Root cause: HostBundler.getRelatedFilesAndContent() calls findContentRelationships() and pulls ALL related contentlets unconditionally; those were then written as .host.xml files regardless of their actual content type, causing either a NotFoundInDbException (ContentType not on receiver) or NPE (host is null) when the receiver processed them

Test plan

• Create a blank site and a Widget with a relationship field pointing to that site
• Push publish the site using the "Only Selected Items" filter
• Confirm push publish completes successfully with no NotFoundInDbException or NPE
• Confirm the site arrives correctly on the receiver
• Run integration test: ./mvnw verify -pl :dotcms-integration -Dcoreit.test.skip=false -Dit.test=ContentHandlerTest

Fixes #34522

🤖 Generated with Claude Code

[claude]

Claude finished @gortiz-dotcms's task in 8m 2s —— View job

---

Code Review

I reviewed the diff against origin/main (the 4 PR commits on top of ca0a0d23). This is a focused, correct root-cause fix. Findings below.

New Issues

No blocking issues found.

Observations (non-blocking)

• 🟡 Medium: HostBundler.java:119-122 — status.addCount() is incremented for contentlets the new guard skips (the loop always calls addCount() after writeFileToDisk, which now returns early for non-Host/non-FileAsset content). The bundler status count will overstate the number of files actually written. This mirrors the pre-existing SYSTEM_HOST early-return at line 318, so behavior is consistent — but if the count is used for reporting/verification it's slightly inaccurate. Minor; not blocking.

• 🟡 Medium: ContentHandler.java:575-586 — the null guard is a sound secondary defense, but note that when host is genuinely expected on the receiver and find() returns null for a transient reason (not just the bundled-non-Host case), this now silently skips updateDefaultHost/updateCache with only a warn. Given the HostBundler fix prevents the actual bad input, this is acceptable defensive handling — just be aware the warning is the only signal if it ever fires for a real host.

Notes on correctness

• HostBundler.java:321 — the guard !isHost() && !isFileAsset() correctly preserves FileAsset bundling (addressed by commit 5938f79a) while excluding Widgets/related content. isHost() (Contentlet.java:1467) resolves by content-type variable, which is the right check. ✅
• The regression test at HostBundlerTest.java:126 is meaningful: without the guard, a related non-Host contentlet reaches writeFileToDisk, gets HOST_EXTENSION appended (line 391-401), and would be picked up by getFileFilter() (.host.xml). The path.contains(relatedContentId) assertion is valid since the identifier appears in the written URI. The test also asserts the Site itself is still bundled. ✅
• The com.dotmarketing.util.FileUtil → com.liferay.util.FileUtil import swap and removal of the unused set static import are consistent with the fixups in 00153e16; no remaining references to the removed symbols. ✅
• @BeforeClass prepare() plus the manual prepare() call inside the hosts() DataProvider is redundant but safe (IntegrationTestInitService.getInstance().init() is idempotent). ✅

Suggestions

• Consider Logger.debug(this, () -> "...") (lambda form) at HostBundler.java:322 to avoid string concatenation when debug logging is disabled — matches the pattern already used elsewhere in the codebase (e.g. ContentHandler.java:563). Non-blocking.

Overall: clean, minimal, and addresses the root cause in the bundler with a reasonable defensive guard in the handler. Good test coverage for the regression.
• branch issue-34522-host-pp-non-host-xml

[claude]

✅ dotCMS Backend Review: no issues found.

[claude]

✅ No Critical or High severity issues found.