Rewrite all time-based checks to use microtime

Food for thought. This adds some reliance upon $cache->created_microtime. It could probably use a little tidying up too.

Wrapped some of the logic (if() statements) to make the lines easier to follow.

Improvements to make the same logic improvements from the cache_flush logic also available on the other scenarios, and added comments explaining the logic.

Some sample code that can be added to a PHP script and then loaded via the webpage to confirm the before & after behavior:

define('DRUPAL_ROOT', getcwd());
include_once DRUPAL_ROOT . '/includes/bootstrap.inc';
drupal_bootstrap(DRUPAL_BOOTSTRAP_FULL);

print("<hr />\n");

$cid = "sidthesloth";
$data = "potato";
$bin = 'cache';

cache_set($cid, $data, $bin, CACHE_TEMPORARY);
print("1. Stored some data in the cache.<br />\n");

$t = cache_get($cid, $bin);
print("2. Cache contains the following:<br />\n");
print("<pre>\n");
print_r($t);
print("</pre>\n");

cache_clear_all("*", $bin, TRUE);
print("3. Caches cleared.<br />\n");

$t = cache_get($cid, $bin);
print("4. The cache object now contains:<br />\n");
print("<pre>\n");
print_r($t);
print("</pre>\n");
print("(That should be a blank value.)<br />\n");

print("<hr />\n");

$cid = "sidthesloth1";
$data = "potato1";
cache_set($cid, $data, $bin, CACHE_TEMPORARY);
print("5. New data stored in the cache.<br />\n");

$t = cache_get($cid, $bin);
print("6. The cache now contains:<br />\n");
print("<pre>\n");
print_r($t);
print("</pre>\n");

print("The cache should return data.<br />\n");

print("<hr />\n");

When the code above is ran against the current -dev release, the following is the results:

1. Stored some data in the cache.

2. Cache contains the following:
stdClass Object
(
    [cid] => sidthesloth
    [data] => potato
    [created] => 1438291623
    [created_microtime] => 1438291623.349
    [flushes] => 0
    [expire] => 1440883622
    [temporary] => 1
)

3. Caches cleared.

4. The cache object now contains:

(That should be a blank value.)

5. New data stored in the cache.

6. The cache now contains:

The cache should return data.

After the patch is applied the output looks like this:

1. Stored some data in the cache.

2. Cache contains the following:
stdClass Object
(
    [cid] => sidthesloth
    [data] => potato
    [created] => 1438291677
    [created_microtime] => 0.90800500
    [flushes] => 0
    [expire] => 1440883674
    [temporary] => 1
)

3. Caches cleared.

4. The cache object now contains:

(That should be a blank value.)

5. New data stored in the cache.

6. The cache now contains:
stdClass Object
(
    [cid] => sidthesloth1
    [data] => potato1
    [created] => 1438291677
    [created_microtime] => 0.93335500
    [flushes] => 0
    [expire] => 1440883674
    [temporary] => 1
)

The cache should return data.

FYI you'll need to clear all sessions after applying the patch because the $_SESSION['cache_flush'] data structure changes.

Working on the tests - they're not all passing yet, so there's some work to be done.

One step closer. I noticed that the MemCacheRealWorldCase tests were failing because the 'language' table wasn't found, so I added 'locale' to the list of enabled modules.

Down to three fails in MemCacheClearCase and one in MemcacheLockFunctionalTest, while MemCacheGetMultipleUnitTest and MemCacheRealWorldCase pass all tests.

FYI the failure in MemcacheLockFunctionalTest is the old "Bootstrap failed in lockInit(), lock_acquire() is not available." problem.

Here's what "drush test-run MemCacheClearCase" has to say for itself:

Starting test MemCacheClearCase.
Cache clear test 75 passes, 3 fails, 0 exceptions, and 9 debug messages
Test MemCacheClearCase->testClearCacheLifetime() failed: Cache item was cleared successfully. in memcache.test on line 504
Test MemCacheClearCase->testClearCacheLifetime() failed: Cache item is not returned once minimum cache lifetime has expired. in memcache.test on line 522
Test MemCacheClearCase->testClearWildcardOnSeparatePages() failed: Cache was properly flushed. in memcache.test on line 701

Some minor code formatting cleanup for the tests. Still just have the four failures.

I left in some debug print() statements in one of the tests, sorry about that.

Reverted the unnecessary changes to the memcache6.test file.

@PieterDC: are you talking about the problem with permissions exported via Features, e.g. #1549608: Cannot import exported Panelizer permissions using Features/defaultconfig if handler cache is stale?

Just to get a baseline of where we are, this is a reroll of the raw changes, it doesn't include the tests.

Was automated testing disabled for the 7.x-1.x branch?