How do I clear Bash's cache of paths to executables?

Question

When I execute a program without specifying the full path to the executable, and Bash must search the directories in $PATH to find the binary, it seems that Bash remembers the path in some sort of cache. For example, I installed a build of Subversion from source to /usr/local, then typed svnsync help at the Bash prompt. Bash located the binary /usr/local/bin/svnsync for "svnsync" and executed it. Then when I deleted the installation of Subversion in /usr/local and re-ran svnsync help, Bash responds:

bash: /usr/local/bin/svnsync: No such file or directory

But, when I start a new instance of Bash, it finds and executes /usr/bin/svnsync.

How do I clear the cache of paths to executables?

Answer 1

bash does cache the full path to a command. You can verify that the command you are trying to execute is hashed with the type command:

$ type svnsync
svnsync is hashed (/usr/local/bin/svnsync)

To clear the entire cache:

$ hash -r

Or just one entry:

$ hash -d svnsync

For additional information, consult help hash and man bash.

Answer 2

There are solutions not mentioned here.

1. You can disable hashing with set +h or set +o hashall

   help set says:

   -h - Remember the location of commands as they are looked up for execution. This is enabled by default.

   hashall - Same as -h

   set -h # enable hashing
   shopt -u checkhash # disable command existence check
   hash -p /some/nonexisting/dir/date date # bind date with /some/nonexisting/dir/date
   date # bash: /some/nonexisting/dir/date: No such file or directory
   set +h
   date # normal date output
   

2. You can check that a command found in the hash table exists before trying to execute it with shopt -s checkhash

   help shopt says:

   checkhash - If set, bash checks that a command found in the hash table exists before trying to execute it. If a hashed command no longer exists, a normal path search is performed.

   set -h # enable hashing
   shopt -u checkhash # disable command existence check
   hash -p /some/nonexisting/dir/date date # bind date with /some/nonexisting/dir/date
   hash -t date # prints /some/nonexisting/dir/date
   date # bash: /some/nonexisting/dir/date: No such file or directory
   shopt -s checkhash # enable command existence check
   date # normal date output
   hash -t date # prints /bin/date
   

3. You can bind NAME with PATH with hash -p PATH NAME or BASH_CMDS[NAME]=PATH:

   shopt -u checkhash # disable command existence check
   hash -p /some/nonexisting/dir/date date
   date # bash: /some/nonexisting/dir/date: No such file or directory
   BASH_CMDS[date]=/bin/date
   date # normal date output
   

4. Magic: PATH="$PATH" performs hash -r

   From variables.c:

   /* What to do just after the PATH variable has changed. */
   void
   sv_path (name)
       char *name;
   {
       /* hash -r */
       phash_flush ();
   }
   

   Try:

   set -h
   hash -r
   date
   hash # prints 1 /bin/date
   PATH="$PATH"
   hash # prints hash: hash table empty
   

Answer 3

To clear just one entry you need a different flag:

hash -d svnsync

The -r flag doesn't take a parameter and will always delete the entire cache.
(At least in bash 3.2.39 on Debian Lenny)

Answer 4

As user johntex has noted in a comment to the answer by user Tobu, the simplest practical action in Bash is to rehash just your program:

hash svnsync

That's all.