About
http://savannah.cern.ch/bugs/?func=detailitem&item_id=4065

-> It is not a bug that the update is sent even if the attachment failed, since we do not refuse the bug posting and the rest of the submitted data is well registered.

-> strlen() is maybe not very efficient for large files, but what else? BTW, on large files, apache/PHP should drop the request by itself.

-> I believe it is on purpose that the filesize test is made after the addslashes(). Otherwise, why not using only filesize(). It is confusing for users, I'm willing to admit it. But file upload is something very sensitive when it comes to webservers, frequently used for exploit. We're forced to rush addslashes() when inserting data in the database to avoid malicious exploits. But I guess someone could act maliciously if we do filesize checks before the addslashes: someone could forge a file to triple the size after the addslashes() call, so he could upload a file way way bigger than the limit that would pass the check.
So in fact, we should probably explain the reason of the refusal more in details, but not change the test.

The applied solution is to feedback the exact size found, after calling addslashes(), in case of refusal.