Hi,
Am 08.01.2014 05:51, schrieb John Hewson:
Yes, it's pre-maven and yes, we should remove it if it's no longer needed.
The patch attached to PDFBOX-1825 does this and is ready to be applied. It
leaves the PDFToImage test disabled so there should be no issues applying the
patch and closing PDFBOX-1825, there are 11 passing tests which are currently
not being run and should be!
I applied the patch in revisions 1557344 and 1557345 but I didn't delete the
TestAll class as it is needed within the ant build.
The PDFToImage test is a special one. It seems not to be deterministic. The
results are different depending on the used environment (JDK, OS), so that
the test fails from time to time without a good reason.
I thought you might say that, I’ll have a think about possible solutions - not
easy. I'll open a new issue for this at some point in the future.
That would be an very useful improvement.
IMO we should add Tilmans diff image idea to illustrate the differences.
-- John
On 5 Jan 2014, at 13:25, Andreas Lehmkuehler <[email protected]> wrote:
Hi,
Am 05.01.2014 19:54, schrieb John Hewson:
Hi All
I’d like to bring https://issues.apache.org/jira/browse/PDFBOX-1825 to the
attention of the mailing list. I’ve noticed that many PDFBox unit tests are
never run due to what appears to be an out-of-date AllTests.java being used
instead of the usual Maven mechanism for running any file name Test**.java
There is a patch attached to the issue to fix this.
Does anyone know the story behind AllTests.java, is it a pre-Maven legacy file
which should have been removed?
Yes, it's pre-maven and yes, we should remove it if it's no longer needed.
Furthermore, what is the goal for regression testing? It would seem that having
more A-B tests for PDFToImage would be a good thing but currently none of these
tests are run.
IMO it's the most obvious, try to detect possible side effects of changes.
Some of them are simple there because they came with a patch to proof its
effectiveness.
The PDFToImage test is a special one. It seems not to be deterministic. The
results are different depending on the used environment (JDK, OS), so that
the test fails from time to time without a good reason.
I've got different results from when using my own jar and using the SNAPSHOT
jar compiled by jenkins. ANd yes, both have the same revision, but compiled
on different machines using different JDKs.
We finally deactivated that test some time ago. Maybe someone has an idea how
to eliminate the dependency on the used enviroment. Otherwise it wouldn't do
much sense to reactivate it.
Cheers
-- John
BR
Andreas Lehmkühler
BR
Andreas Lehmkühler
