Changes

Initial commit.

Add the particulars to the info file.

Add the drush command file.

Add the configuration for supervisor.

Add the bash script to run the drush command.

Add the module file.

Surface the feature to Aegir.

Add a simple readme.

Remove an extra blank line.

Add a settings form.

Add the hook_menu().

Add a description to the admin form.

Add a enable and disable hook to turn off the hosting tasks dispatch.

Add more detail to the delay options.

Beef up the README

Add some feedback that the queue runner is running.

Add a note about trouble shooting.

Add some more logging to the drush command.

Make sure the description of the form shows up.

Reduce the sleepy time, we weren't feeling responsive any more.

Add line breaks to the README.

restart the queue runner automatically after the timeout, see #1189696

Issue #1189696 by Steven Jones: Fixed Running without supervisord leaks memory.

make the process lifetime customizable

Issue #1189768 by anarcat, Steven Jones: Added customizable process lifetime.

make delays translatable

simplify the fork call and rename the function to be clearer

rewrite the pcntl_exec() stuff to work with different command names and respect provided options

always restart the dispatcher

whitespace and comments

log to watchdog so frontend has feedback of dispatcher

add one minute process lifetime

pass process environment to child task correctly

change the feedback to remove some duplicate and useless messages

add init script to start our shiny daemon

document the autorestart feature and the init.d file

word wrap

handle SIGHUP as a restart properly

now it interrupts child tasks, which isn't good, but can be improved

wait for the child tasks before restarting after signal

handle the case where ENV is not set

simplify signal handling code

we do not need to fork, as we do not need the PID to wait for it,
thanks to pcntl_wait().

make init script executable

remove unused variable

handle stopping gracefully, by waiting for our poor children

better log messages for signals

use unset instead of assigning to null to try to free memory

load more than one task at a time

this reduces latency (one less sleep(1) call), the number of SQL
requests (one request per N tasks, where N is the default argument of
hosting_get_new_tasks()) and probably use less memory because we don't
do the wider loop all the time

this means, however, we need to check the time in the inner loop too

Fix up the auto restart functionality and remove forking.

Revert "make delays translatable" format plural is translatable.

This reverts commit 5d13147501a43630f1abb6e2fe87009a67abc919.

Tidy up.

Cleanup to ensure we can use PCNTL.

rename the readme to follow conventions

clarify why we pause twice

Argh that merge was wrong, fix it!

Clean up some docs.

Fix formatting in the README.

fixed status check to match the process as it actually runs

import hosting_queue_runner as a submodule in 2.x

1548490 by grobot - fix hosting-task call for drush 5

rename the hosting-queue-runner to hosting-queued

a few reasons:

 1. it's shorter
 2. it's really a daemon
 3. it will look better in init.d/

1613376 by greg.1.anderson: fix invoke so that the alias is passed properly