Corpus structure

    1. archive/$version/forward_incompat/$type
    • Checkout a clean repo (best not to do this where you normally work):
    1. git clone ceph.git
    2. cd ceph
    3. git submodule update --init --recursive
    • Build with flag to dump objects to /tmp/foo:
    1. rm -rf /tmp/foo ; mkdir /tmp/foo
    2. do_cmake.sh -DCMAKE_CXX_FLAGS="-DENCODE_DUMP_PATH=/tmp/foo"
    3. cd build
    4. make
    • Start via vstart:
    • Use as much functionality of the cluster as you can, to exercise as many object encoder methods as possible:
    1. bin/ceph osd pool create mypool
    2. bin/rados -p mypool bench 10 write -b 123
    3. bin/ceph osd in 0
    4. bin/init-ceph restart osd.1
    5. for f in ../qa/workunits/cls/*.sh ; do PATH="bin:$PATH" $f ; done
    6. PATH="bin:$PATH" ../qa/workunits/rados/test.sh
    7. bin/ceph_test_librbd
    8. bin/ceph_test_libcephfs
    9. bin/init-ceph restart mds.a
    10. ../qa/workunits/rgw/run-s3tests.sh
    • Stop:
    1. ../src/stop.sh
    • Import the corpus (this will take a few minutes):
    1. ../src/test/encoding/import.sh /tmp/foo `bin/ceph-dencoder version` ../ceph-object-corpus/archive
    2. ../src/test/encoding/import-generated.sh ../ceph-object-corpus/archive
    • Prune it! There will be a bazillion copies of various objects, and we only want a representative sample.:
    • Verify the tests pass:
      • Commit it to the corpus repo and push:
      1. pushd ../ceph-object-corpus
      2. git checkout -b wip-new
      3. git add archive/`../build/bin/ceph-dencoder version`
      4. git commit -m `../build/bin/ceph-dencoder version`
      5. git remote add cc git@github.com:ceph/ceph-object-corpus.git
      6. git push cc wip-new
      7. popd
      • Go test it out:
      1. cd my/regular/tree
      2. cd ceph-object-corpus
      3. git fetch origin
      4. git checkout wip-new
      5. cd ../build
      6. ctest -R readable.sh