Note that there are some explanatory texts on larger screens.

plurals
  1. USpeabody
    primarykey
    data
    text
    plurals
    1. CODepends what you're caching. If the information cached is private to a user, a hash collision could mean leaking private data (which I would call an earth shattering bug). Granted there are hashes that make hash collision probability effectively nil, but crc32 isn't one of them. It only takes slightly over 77000 values for the odds of a hash collision to be 1 in 2 (http://preshing.com/20110504/hash-collision-probabilities). Using a better hash function for the key helps, but every hash table implementation I know of accounts for collisions. It's not hard; just store and check the key.
      singulars
    2. COI thought about this myself. I just always wondered if hashing to md5 was somehow reducing the security. Another way I thought of is that you could have an indicator tied to the account which indicates if the password is old or new. If new, hash to md5 then bcrypt. If new, just bcrypt. This, would allow you to convert the existing database completely so you aren't waiting on users to convert their old passwords.
      singulars
    3. COThis should be the accepted answer. It's precisely what the error message is referring to. In my case when I cloned I had used a FQDN, but on my new machine I had only logged in using the short local domain name. I had to login via putty or plink as the FQDN to cache the key for the host name on the origin. It can help to cross check the hostname being used as the remote using "git remote -v".
      singulars
 

Querying!

 
Guidance

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload