test(release-smoke): cover the render-pptx and render-docx artifacts#440
Merged
Conversation
The consumer smoke harness is the authoritative "a real user can install this" gate, but it only ever exercised the PDF stack. Both POI backends could ship with a broken pom, a missing ServiceLoader resource, or an unpublished coordinate and the harness would still report green — and render-pptx is the headline artifact of this release. Add s7 (core + render-pptx) and s8 (core + render-docx + render-pdf). Both inspect the emitted OPC package with java.util.zip rather than Apache POI: the point is to prove the published artifacts work for a consumer who installed nothing else, so a scenario must not pull a parsing library of its own to make its assertions pass. s7 asserts more than a valid ZIP header, which an empty file would satisfy: exactly one slide part, the 16:9 slide dimensions in EMU, editable text runs, native shapes, and no picture at all — the failure mode that would mean the backend silently rasterised everything. It also resolves the provider by format. It needs 2.1.0 or later, since neither the fixed-layout backend nor DocumentPageSize.SLIDE_16_9 exists in 2.0.0. s8 carries render-pdf because the combination does not work without it: opening a DocumentSession resolves a FontMetricsProvider, and render-pdf is the only artifact that publishes one. render-docx declares it at test scope for its own suite, so the module's tests pass while a consumer installing core + render-docx alone fails at create() before any export runs. render-docx/README.md now states that requirement. Bump the harness's default version from the release script. Four files hard-coded the version to smoke-test when none is passed, and the script never touched them, so a post-release run that accepted the prefilled default re-verified the *previous* release and reported green — the one failure mode a release gate must not have. Each pattern is anchored to its own construct, and the bump sits on the final-release path so a release candidate correctly leaves it alone. Verified: full reactor clean verify green (1518 tests, 0 failures). Both scenarios run green against a locally installed 2.1.0-SNAPSHOT; s8 fails without render-pdf, which is how the packaging gap surfaced. All three cut-release dry-run modes pass — the full cut bumps all six sites and stages the four files, the RC path performs zero smoke bumps. The <p:pic> assertion was checked against a generated deck that does contain a picture, so it is not vacuously true.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
The consumer smoke harness is the authoritative "a real user can install this" gate, but it only ever exercised the PDF stack (s1–s6). Both POI backends could ship with a broken pom, a missing
META-INF/servicesresource, or an unpublished coordinate and the harness would still report green — andgraph-compose-render-pptxis the headline artifact of 2.1.0.Separately, four files hard-coded the version to smoke-test when none is passed, and the release script never touched them. A post-release run that accepted the prefilled default re-verified the previous release and reported green — the one failure mode a release gate must not have.
What changed
s7 —
graph-compose-core+graph-compose-render-pptx. Asserts more than a valid ZIP header, which an empty file would satisfy: exactly one slide part, the 16:9 slide dimensions in EMU, editable text runs, native shapes, and no picture at all — the failure mode that would mean the backend silently rasterised everything. Also resolves the provider by format. Requires 2.1.0+, since neither the fixed-layout backend norDocumentPageSize.SLIDE_16_9exists in 2.0.0.s8 —
graph-compose-core+graph-compose-render-docx+graph-compose-render-pdf. The PDF backend is in the set because the combination does not work without it: opening aDocumentSessionresolves aFontMetricsProvider, and render-pdf is the only artifact that publishes one.render-docx/pom.xmldeclares it at test scope for its own suite, so the module's tests pass while a consumer installing core + render-docx alone fails atcreate()withMissingBackendExceptionbefore any export runs. This PR documents the requirement inrender-docx/README.md; the underlying packaging fix belongs to the backend-neutral font-measurement extraction and is filed separately.Both scenarios inspect the emitted OPC package with
java.util.ziprather than Apache POI. The point is to prove the published artifacts work for a consumer who installed nothing else, so a scenario must not pull a parsing library of its own to make its assertions pass.Harness default version is now bumped by
cut-release.ps1. Six sites acrossrun.sh,run.ps1,release-smoke.ymland the harness README, each pattern anchored to its own construct so a bump cannot smear across unrelated version strings. It sits on the final-release path, so a release candidate correctly leaves it alone.Verification
./mvnw -B -ntp clean verify— BUILD SUCCESS, 1518 tests, 0 failures.2.1.0-SNAPSHOT. s8 fails without render-pdf — that is how the packaging gap surfaced, not a guess.cut-release.ps1dry-run modes pass: the full cut bumps all six sites and stages the four files;-PostReleaseOnlyis clean; the RC path performs zero smoke bumps.<p:pic>assertion was checked against a generated deck that does contain a picture (maven-banner.pptx): POI emits the literal<p:pic>, and<p:sp>does not collide with<p:spTree/<p:spPr>. Neither assertion is vacuously true.