NullPointerException when generating PDF report

[eclipse-birt]

NullPointerException when generating PDF report

[hvbtup]

I can say that this NPE does not occur with the BIRT version from https://github.com/hvbtup/birt which I built on 20200804 (where I originally fixed this and took the patch from).
So the reason must be in the differences between this version and the current eclipse/birt.

[hvbargen]

Debugging this is quite hard and time-consuming, particularly because I don't know how I could debug without calling mvn package (takes ~30min here), extract the 240 MB ZIP all-in-one, then start it and run the test report.
What I could find out so far by adding some debug messages in drawBorder is:
In the first call to PageDeviceRender::drawBorder, the expression tb.rowBorders.get( tb.tableLRY ) returns null, which causes the NPE, while String.valueOf(tb.rowBoders) looks like this:
{-31015=org.eclipse.birt.report.engine.layout.emitter.TableBorder$Border@604e6493, -34015=org.eclipse.birt.report.engine.layout.emitter.TableBorder$Border@54e7f03}

[hvbargen]

The source code (eg TableBorder.java) is an undocumented mess, unfortunately, for example: It is normal that the keys in the map are negative? I don't know. One can only guess the purpose of most classes. I'll create a toString() method for TableBorder and see if this helps.
BTW if I add if-statements to handle the null border case, then the report doesn't fail, but the page-breaking is different: The yellow text is on the first page then instead of causing a page-break.

[hvbargen]

I narrowed this down a bit by testing the report with several older BIRT releases.
With BIRT 4.3.0, 4.4.0, 4.5.0, 4.6.0 the rendered PDF is looking wrong (like I said https://bugs.eclipse.org/bugs/show_bug.cgi?id=562873).
With BIRT 4.7.0 I get the NPE.
So obviously this bug was introduced with BIRT 4.7.0.
That means I only have to look at the commits from ~ 1 year between june 2016 and june 22, 2017.

[hvbargen]

What I wanted to do is binary searching for the exact commit which introduced the NPE bug.
But I cannot build older versions myself.
For example, if I call

cd birt
git checkout 7d46aaaf5730dc312915c84492f60ce62eb5fbbe
cd ..
build (which sets up environment vars and calls mvn package -DskipTests)

then this results in a lot of errors like

[FATAL] Non-resolvable parent POM for org.eclipse.birt.build:org.eclipse.birt.build-parent:0.0.1-SNAPSHOT: Could not find artifact org.eclipse.birt:org.eclipse.birt-parent:pom:0.0.1-SNAPSHOT and 'parent.relativePath' points at wrong local POM @ line 4, column 10

Any ideas what to do here?
What I wanted to do is binary searching for the exact commit which introduced the NPE bug.

[wimjongman]

you have to build in the birt directory.

[hvbtup]

I know, and that's what my build.cmd script (located inside parent directory of birt) does, and the script is working fine with the current head:

setlocal
set JAVA_HOME=D:\prog\jdk8
set PATH=D:\prog\apache-maven-3.6.3\bin;%JAVA_HOME%\bin;%PATH%
pushd birt
mvn clean package -DskipTests
popd
endlocal

[wimjongman]

Older versions have to be run with JDK8.

[hvbtup]

I know, and my script takes care of that. Maybe some GIT-ignored artifacts (from a build with the head revision) are still around and cause trouble. I'll try from scratch this weekend.

[wimjongman]

Can you run 'mvn clean package -DskipTests' manually from the root dir?

[hvbtup]

Same result.
I also tried with a clean directory (I deleted everything except .git, then git checkout 7d46aaa, then git restore, then build.
Still same result.
I don't know about POM, P2 etc, but maybe this is wrong thrEoughout the whole project in this earlier version:

birt\model\pom.xml says:

<project xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd" xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
<modelVersion>4.0.0</modelVersion>

<parent>
	<groupId>org.eclipse.birt</groupId>
	<artifactId>org.eclipse.birt-parent</artifactId>
	<version>0.0.1-SNAPSHOT</version>
	<relativePath>../</relativePath>
</parent>
...

And birt\pom.xml says:

<?xml version="1.0" encoding="UTF-8"?>
<project xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd" xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
    <modelVersion>4.0.0</modelVersion>
    <groupId>org.eclipse.birt</groupId>
    <artifactId>org.eclipse.birt-parent</artifactId>
    <version>4.6.0-SNAPSHOT</version>
    <packaging>pom</packaging>
...

Isn't this a version mismatch 4.6.0 <-> 0.0.1 ?

[wimjongman]

Yes, it is a mismatch. They must be the same.

[hvbargen]

That was a good tip, but this now results in another build error.

I replaced <version>0.0.1-SNAPSHOT</version> with <version>4.6.0-SNAPSHOT</version> in all pom.xml files with the help of a good editor, then tried to build again.

Now I have

[ERROR] Internal error: java.lang.RuntimeException: Failed to load p2 repository with ID 'eclipse' from location http://download.eclipse.org/eclipse/updates/4.6milestones/: No repository found at http://download.eclipse.org/eclipse/updates/4.6milestones. -> [Help 1]
org.apache.maven.InternalErrorException: Internal error: java.lang.RuntimeException: Failed to load p2 repository with ID 'eclipse' from location http://download.eclipse.org/eclipse/updates/4.6milestones/

It seems as if the whole build process was (or still is) broken somehow.
It shouldn't be that damn hard to compile BIRT based on an older commit.

[wimjongman]

You are talking 2016. I'm surprised if any of the old repositories still exist. If you click that link you will see that it has moved to an archive location.

[hvbargen]

I could narrow it down to commit f84d0f8 "Do not use specified height but actual height if row is empty to avoid endless page break".

For a test, I reverted the changes from commit f84d0f8 in file org\eclipse\birt\report\engine\nLayout\area\impl\TableLayout.java.

Now the test report is working as expected - without a NullPointerException. The second page looks like this.

[hvbargen]

I only wonder why the current BIRT head reacted with a NPE whereas when I fixed the very same bug in my fork (see hvbtup@b8de757), BIRT reacted with a wrong layout.
But anyway...
I'll create a PR soon.

[hvbargen]

I tested that my fix (reverting an old commit) worked with the with the example report from https://bugs.eclipse.org/bugs/show_bug.cgi?id=562873.

But after I merged the latest changes, created a new branch "pr_issue_636" and performed the same little change again, I can no longer use the generated report-all-in-one IDE.

HTTP ERROR 503
Problem accessing /viewer/preview. Reason:
Service Unavailable

It seems like I'm hitting issue #621 or #620 (but I'm on Windows).

I'll create a PR nevertheless.

[hvbargen]

Created PR #644

[wimjongman]

Thanks, @hvbargen!