GitLab

There is a bug in PMD, see https://github.com/pmd/pmd/issues/872
This is likely properly fixed in 6.1.0.
But for 5.7.0, this custom fix should work in DaCapo's case.

 - changed checking from stderr to stdout;
 - updated checksum;
 - use `braces.xml` and `text` format to produce some output on the
`xsmall` size;
 - patch a problematic code in PMD `AbstractRenderer.java` which closes
output stream after flushing. This causes null pointer error when DaCapo
tries to close TeePrintStream again.

`build.xml`:
  - add library dependency -- saxon;
  - skip testing when building PMD so to build faster;
  - correct inclusion paths for Java source files and rulesets;

`pmd.cnf`:
  - update version and website
  - update rule set paths and remove non existing rule sets:

    - favorites.xml
    - newrules.xml
    - scratchpad.xml

PMD 5.7.0 supports passing in file lists; thus `prepended_filelist()` is
added to generate a new file list with scratch directory prepended to
each entry in the list. The original file list is renamed with an
`.orig` suffix.

`prepare()` is modified to conform to the new commandline interface. In
particular, flags are added and language and version are specified.

Also sets a system property so that PMD doesn't just exit after run.

PMD 5.7.0 use Maven as its build system. The `build.xml` for the
benchmark has been changed to call Maven.
Also, it requires Java JDK toolchain to be configured for Maven. A
shell script has been added to generate the toolchain.xml file.

Note that during the building process, PMD requires both 1.7 and 1.8
toolchain to be configured. A clean solution would require the JDK7 to
be installed on the platfor along with JDK8/9. This requirement can be
annoying. Therefore the generated toolchain.xml file
has a hack such that if the platform java version is >= 1.8, it will
generate two entries (1.7 and the actual version) that point to the same
JDK directory.

Also the download link for the release has been changed to GitHub.

Create new benchmark "LusearchFix" which is identical to lusearch (except name)

Add patch file for lusearchFix for patching lucene. The changing will make lusearchFix has differenct performances compare with lusearch

Changed the name from lusearchFix to lusearch-fix

Added codes for letting lusearch-fix get data from lusearch, eliminating duplications

Change the path of the LICENSE

Change the path of README.md

Fix the broken link problem for h2 and junit

Fix the broken link for janino package

Update the patch file for tomcat for fixing one broken link

Add the patch file for libs/commons-logging to fix the broken link for downloaded package

Fix the broken link of jstl and update the checksum for jstl

Delete the extra space in md5 checksum file for jstl

Deleting the echo in libs/common.xml

fix: assuming DaCapo is managed by git, properly extract commit hash from git

remove extra space

no fail on error

fix: correctly point to jar file so to extract version info

Fixed the error in tomcat and pmd

fixed typo

In particular, removed DaCapo-specific tomcat classes from main dacapo.jar.

Addressed issue #3049886 (jython generates NPE with eager compilation) by updating to latest release (Jython 2.5.2).

Addressed issue #3056019 and related issues (fixed multiple system properties during actual benchmark iteration).

Added a template method with which to augment the system properties during the actual benchmark iteration.

Added a template method with which to augment the system properties during the actual benchmark iteration.