pom_dlineage.xml: drop the private <parent>, and build it in CI so this stops hiding

pom_dlineage.xml declared <parent>gudusoft:gsp_java:1.0-SNAPSHOT</parent>, the
private library reactor, which is published nowhere. On any machine that had
not installed that POM, Maven stopped before reading the file's own contents:

    Non-resolvable parent POM for gudusoft:gsp_demo_java_dlineage:1.0-SNAPSHOT:
    Could not find artifact gudusoft:gsp_java:pom:1.0-SNAPSHOT

The root pom.xml was cut loose from that same parent when this repository was
made buildable outside Gudu, and carries a comment saying so. This file was
missed, and it looked fine to everyone here because we all have the library
checked out beside it.

Checked what would be lost before removing it: two properties
(maven.compiler.source/target 1.8, both already declared here), an empty
<dependencies>, and a single maven-antrun-plugin marked <inherited>false</inherited>
that deletes .pas files out of the library's parser folder. No
dependencyManagement, pluginManagement, distributionManagement or repositories.
Nothing needed.

Verified against a completely empty local repository, which is the only way to
see this: the build succeeds, creates no gudusoft/ group directory at all, and
its classpath takes com.gudusoft:gsqlparser:4.1.9 from the public Maven repo,
with lib/ contributing only simple-xml, fastjson and expr4j. The packaged jar
runs and emits lineage XML.

Root cause of it surviving this long: no workflow built this file. Both now do,
and they also assert it left the root build's target/classes alone, since
sharing ${project.basedir} with a single-source-file build is what made it wipe
the root build in issue #39.

That is the second breakage found in this one file, both invisible to us and
both fatal for a plain clone. Noted in CLAUDE.md that anything CI does not run
is load-bearing only by luck.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012qPRpoD8exYRrUmbfXXWXj

Author: 2 people

.github/workflows/build.yml

@@ -75,6 +75,32 @@ jobs:
           echo "$out"
           grep -q "syntax errors: 0" <<<"$out"
 
+      # pom_dlineage.xml is a second, separate build in this same directory, and
+      # nothing used to run it. It declared a <parent> of gudusoft:gsp_java,
+      # the private library reactor, so it only built for someone who had that
+      # POM installed locally; a plain clone got "Non-resolvable parent POM"
+      # before Maven read any of its own contents. Building it here is what
+      # stops that class of thing hiding again.
+      #
+      # It also has to leave the root build alone: it declares a single source
+      # file, so sharing target/ with the root build makes the compiler plugin's
+      # incremental cleanup delete every class the root build just produced
+      # (issue #39). That is why it writes to target-dlineage/, and why the
+      # count check below is worth the two lines.
+      - name: Build the separate dlineage POM, and check it leaves target/ alone
+        shell: bash
+        run: |
+          set -euo pipefail
+          before=$(find target/classes -name '*.class' | wc -l)
+          mvn -B -f pom_dlineage.xml package -DskipTests
+          after=$(find target/classes -name '*.class' | wc -l)
+          echo "root classes before=$before after=$after"
+          if [ "$before" != "$after" ]; then
+            echo "::error::pom_dlineage.xml changed the root build's target/classes ($before -> $after); see issue #39"
+            exit 1
+          fi
+          ls -l target-dlineage/*.jar
+
   # The .bat scripts are the original Windows, no-Maven workflow: edit
   # setenv\setenv.bat, cd into a demo folder, run compile_<demo>.bat then
   # run_<demo>.bat. They had been stale for years -- compiling

.github/workflows/nightly.yml

@@ -83,6 +83,24 @@ jobs:
       - name: Check the results
         run: .github/scripts/check-test-results.sh
 
+
+      # A second, separate build in the same directory that nothing used to run.
+      # It carried a <parent> pointing at the private gudusoft:gsp_java reactor,
+      # so it only built where that POM happened to be installed. Also asserts it
+      # leaves the root build's target/classes alone (issue #39).
+      - name: Build the separate dlineage POM
+        shell: bash
+        run: |
+          set -euo pipefail
+          before=$(find target/classes -name '*.class' | wc -l)
+          mvn -B -f pom_dlineage.xml package -DskipTests
+          after=$(find target/classes -name '*.class' | wc -l)
+          if [ "$before" != "$after" ]; then
+            echo "::error::pom_dlineage.xml changed the root build's target/classes ($before -> $after); see issue #39"
+            exit 1
+          fi
+          ls -l target-dlineage/*.jar
+
       - name: Every demo starts
         run: .github/scripts/run-all-demos.sh --timeout 90
 
@@ -153,6 +171,24 @@ jobs:
       - name: Check the results
         run: .github/scripts/check-test-results.sh
 
+
+      # A second, separate build in the same directory that nothing used to run.
+      # It carried a <parent> pointing at the private gudusoft:gsp_java reactor,
+      # so it only built where that POM happened to be installed. Also asserts it
+      # leaves the root build's target/classes alone (issue #39).
+      - name: Build the separate dlineage POM against the newest release
+        shell: bash
+        run: |
+          set -euo pipefail
+          before=$(find target/classes -name '*.class' | wc -l)
+          mvn -B -f pom_dlineage.xml package -DskipTests -Dgsp.core.version=${{ steps.ver.outputs.version }}
+          after=$(find target/classes -name '*.class' | wc -l)
+          if [ "$before" != "$after" ]; then
+            echo "::error::pom_dlineage.xml changed the root build's target/classes ($before -> $after); see issue #39"
+            exit 1
+          fi
+          ls -l target-dlineage/*.jar
+
       - name: Every demo starts
         env:
           MVN_ARGS: -Dgsp.core.version=${{ steps.ver.outputs.version }}

README.md

@@ -695,6 +695,29 @@ moved on to `getOption().setTraceTablePosition(...)` and
 3.1.1.0 jar has. It now resolves the same parser the root build does, and
 builds and runs.
 
+**It then failed a second way, for anyone but us.** It kept a `<parent>` of
+`gudusoft:gsp_java:1.0-SNAPSHOT`, the private library reactor, which is not
+published anywhere. On a machine that had never installed that POM, Maven
+stopped before reading the file at all:
+
+```
+Non-resolvable parent POM for gudusoft:gsp_demo_java_dlineage:1.0-SNAPSHOT:
+Could not find artifact gudusoft:gsp_java:pom:1.0-SNAPSHOT
+```
+
+The root `pom.xml` was cut loose from that parent when this repository was made
+buildable outside Gudu; this file was missed, and it looked fine to everyone who
+had the library checked out locally. The parent contributed nothing it needed:
+two compiler properties it already declares, an empty `<dependencies>`, and one
+`maven-antrun-plugin` marked `<inherited>false</inherited>`. It is gone, and the
+build now resolves `com.gudusoft:gsqlparser` from Gudu's public Maven repository
+and nothing else — verified against a completely empty local repository, where
+no `gudusoft/` group directory is created at all.
+
+Both workflows now build it, and assert it left the root build's
+`target/classes` untouched. Nothing ran it before, which is why two separate
+breakages could sit in it unnoticed.
+
 ## master and dev branches
 
 `master` is updated when a new GSP version is released on

pom_dlineage.xml

@@ -4,11 +4,24 @@
          xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
   <modelVersion>4.0.0</modelVersion>
 
-  <parent>
-        <groupId>gudusoft</groupId>
-        <artifactId>gsp_java</artifactId>
-        <version>1.0-SNAPSHOT</version>
-    </parent>
+  <!-- No <parent>, for the same reason the root pom.xml has none: this used to
+       inherit from gudusoft:gsp_java:1.0-SNAPSHOT, the private library reactor,
+       which is not published anywhere. Anyone without that POM installed in
+       their local repository got
+
+           Non-resolvable parent POM ... Could not find artifact
+           gudusoft:gsp_java:pom:1.0-SNAPSHOT
+
+       before Maven even looked at this file's own contents. The root pom.xml
+       was cut loose when the repository was made buildable outside Gudu; this
+       one was missed, and nothing noticed because no workflow builds it.
+
+       The parent contributed nothing that is needed here: its only properties
+       are maven.compiler.source/target 1.8, both declared below; its
+       <dependencies> is empty; and its one plugin is maven-antrun-plugin marked
+       <inherited>false</inherited>, cleaning .pas files out of the library's
+       parser folder. There is no dependencyManagement, pluginManagement,
+       distributionManagement or repositories section to lose. -->
 
   <groupId>gudusoft</groupId>
   <artifactId>gsp_demo_java_dlineage</artifactId>