* Start assembling a cmake setup with a mix of Kudu and dynd-python snippets and cmake modules

* Build native.pyx stub and libpandas.so
* In place cmake build + rpath on linux, at least
* Library scaffolding and build infrastructure for libpandas native core
  experiments to evaluate replacing Series and DataFrame internals and custom
  dtypes with a consistent set of wrapper types, array, and table data
  structures. Incorporates several bits of BSD / MIT / Apache-licensed code and
  various copyright notices have been added to LICENSES.
* Do not include pandas/
* Array API stubs
* Import bunch of utility code from Arrow, refactoring

Author: wesm

.travis.yml

@@ -1,4 +1,16 @@
-sudo: false
+sudo: required
+dist: trusty
+addons:
+  apt:
+    sources:
+    - ubuntu-toolchain-r-test
+    - kalakris-cmake
+    packages:
+    - gcc-4.9
+    - g++-4.9
+    - cmake
+    - valgrind
+
 language: python
 
 # To turn off cached miniconda, cython files and compiler cache comment out the
@@ -250,6 +262,8 @@ before_install:
   - source ci/travis_process_gbq_encryption.sh
   - echo $VIRTUAL_ENV
   - export PATH="$HOME/miniconda/bin:$PATH"
+  - export APT_ARGS="-y"
+  - sudo apt-get install ccache
   - df -h
   - date
   - pwd
@@ -272,6 +286,8 @@ install:
 before_script:
   - source activate pandas && pip install codecov
   - ci/install_db.sh
+  - # mysql -e 'create database pandas_nosetest;'
+  - # psql -c 'create database pandas_nosetest;' -U postgres
 
 script:
   - echo "script start"