Permalink
Commits on Dec 18, 2012
  1. Merge pull request #78 from kyleconroy/separation

    kyleconroy committed Dec 18, 2012
    Separation
  2. No need to change Makefile

    kyleconroy committed Dec 18, 2012
Commits on Dec 17, 2012
  1. Fix merge conflicts

    kyleconroy committed Dec 17, 2012
  2. Merge pull request #68 from msabramo/testing

    kyleconroy committed Dec 17, 2012
    Testing - nosetests, tox, and Travis CI
Commits on Dec 12, 2012
  1. Merge pull request #76 from kelvl/master

    progrium committed Dec 12, 2012
    Fixes for SUID/SGID and SIGTERMs
Commits on Aug 21, 2012
  1. Added pwd and grp to imports for runner so setuid and setgid works

    kelvl committed Aug 21, 2012
    Changed the order in which setuid and setgid is ran so that the process doesn't lose privilege to setgid after setuid completes
    
    Removed recursive child.serve_forever() from serve_forever(),
    this was creating a bug where a child service's serve_forever() blocks, because the loop runs sequentially, the child's siblings' serve_forever() doesn't yet run.
    
    When stop() is issued on the parent, the child's serve_forever() finishes, and the sibling's serve_forever() is called, which starts the sibling again, leading to an inconsistent state and the process never dies. This leads to the parent ginkgo service to not respond correctly to SIGTERMs
    
    I believe the serve_forever() loop is redundant, because self.state.wait('stopped') will have already be waiting for all child services to stop. The parent will only transition to stopped after all child services have stopped.
Commits on Jun 23, 2012
Commits on Jun 22, 2012
Commits on Jun 21, 2012
  1. fixing weird amend

    progrium committed Jun 21, 2012
  2. Wrote GlobalContext mixin for use with Process and Config to make the…

    progrium committed May 18, 2012
    …m context managers that will set their global ginkgo singleton while in that context. Also, GlobalContext has tests.
Commits on May 30, 2012
  1. add serve_forever back to basicservice

    Alan Shreve committed May 30, 2012
Commits on May 24, 2012
  1. remove gevent dependency

    Alan Shreve committed May 24, 2012
  2. fix threading's do_stop function to approximate the functionality del…

    Alan Shreve committed May 24, 2012
    …ivered by the other async modules.
Commits on May 23, 2012
  1. move around signal imports

    Alan Shreve committed May 23, 2012
  2. move the most common and default implementations of the signal/init f…

    Alan Shreve committed May 23, 2012
    …unctions into AbstractAsyncManager
  3. Merging in the eventlet branch and tweaking it a bit to match the sli…

    Alan Shreve committed May 23, 2012
    …ghtly modified async interface.
  4. Implement an async manager for threads. Make 'async' a configuration …

    Alan Shreve committed May 23, 2012
    …option that allows you to set the default async manager for services. Improved help messages when choosing invalid async managers. Removed serve_forever and its use in runner to fix bugs which would sometimes cause services to be started as they were stopping.
Commits on May 18, 2012
  1. Wrote GlobalContext mixin for use with Process and Config to make the…

    progrium committed May 18, 2012
    …m context managers that will set their global ginkgo singleton while in that context. Also, GlobalContext has tests.
Commits on May 15, 2012
  1. calling group on a Config object now returns None if the path given i…

    progrium committed May 15, 2012
    …s actually a value and not a group. And Groups now also act as read-only dicts using the collections.Mapping mixin/base.
  2. changed default logfile from /tmp/ServiceName.log to ~/.ServiceName.l…

    progrium committed May 15, 2012
    …og plus some other refactoring
  3. made subclass of Process called DaemonProcess to avoid repeating if s…

    progrium committed May 15, 2012
    …tatements. prepare_app was renamed to setup_process and takes a daemonize option. daemon is no longer configurable. when you use ginkgoctl, it will daemonize. when you use ginkgo it will not, unless you give it the -d argument. pidfile is only created when you daemonize, so is handled only in DaemonProcess. lastly, pidfile is stored as a dot file in your home directory instead of in /tmp by default.
  4. Add .travis.yml

    msabramo committed May 15, 2012
Commits on May 11, 2012
  1. removing examples. they're so old they're starting to become more har…

    progrium committed May 11, 2012
    …mful than useful. new examples should be written
  2. added pre-wrapped gevent servers. threw in a simple StreamClient serv…

    progrium committed May 11, 2012
    …ice. ServerWrapper is not deprecated
  3. added module docstrings to all modules. maybe not the best, but a sta…

    progrium committed May 11, 2012
    …rt. also some minor refactorings
Commits on May 10, 2012
  1. added module docstring, refactored one-off functions used by daemoniz…

    progrium committed May 10, 2012
    …e, and removed unused imports
  2. raise exception when configuration loading fails, but catch it and lo…

    progrium committed May 10, 2012
    …g as warning in process reload handler. this means if you use ginkgoctl to reload, it will fail before it even reloads. if you do a manual SIGHUP, it will log a warning that config failed to reload and will for now continue on with reloading of children
  3. as it turns out, we want reload to behave the same way as start in re…

    progrium committed May 10, 2012
    …gards to using start_before