Rebuild permissions generates Fatal error: spl_autoload(): Class Database could not be loaded in database.inc on line 2412

Oops, wrong module..

xjm, I'm using mysql. There are no (!) messages in the log. I am not familiar with backtracing - my IDE consists of Editpad/notepad..

OK, it might take a while, but i'll be back with more info.

I am not sure if this error belongs to this module.
I have the same error every time I generate a Views display, which only differs slightly from another display, that is working fine....

Above error happens also when only the module Node access user reference is used, so this seems to be a core issue.
This is the line of code:

function db_merge($table, array $options = array()) {
  if (empty($options['target']) || $options['target'] == 'slave') {
    $options['target'] = 'default';
  }
  return Database::getConnection($options['target'])->merge($table, $options);
}

More info: The function db_merge is calle 1 time for table cache and 20 times for table cache_field .
So I cleared my caches (thus the table), but during 'Rebuild content access', the table again grows happily to 14,099 lines, until Rebuilding stalls at 95%. Then, in db_merge, the error occurs.

Is cache_field too big? can I avoid caching during Rebuild permissions?

During the execution of Rebuild permissions, I executed cache_clear_all('*', 'cache_field', TRUE); using devel/php page. At 95%, there were only around 500 lines in the table.
Same result.. :-(

Can't find a good spot for backtracing..
I thought this might be another symptom of the same problem: #1375748: Too many node_access-records
But apparently it is normal behaviour that every node gest 1 line in node_access table.

xjm, thanks for your help. I have a successful rebuild-run. I am not sure about the cause, but I used your hint about the corrupt data. This is what I did:
- I learned that node_access-rebuild rebuilds starting with lowest nid, in chunks of 20 nodes.
- So I checked the highest nid in node_access table;
- edited and saved the next 25 nodes; (none of them seamed erroneous)
- waited until non-peak hours, to make sure server has lowest capacity-needs;
- flushed all caches;
- did a Rebuild --> Success!

I will have to do some more runs to be sure that this was the root cause, but for now I am happy.

I think current issue-status is OK for the moment...

FYI, Today I tried again a rebuild.
First time, I got the same Fatal error: spl_autoload(). I did a cache-flush and after that, the rebuild was OK. Strange... just a normal short of memory??

.