Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


  • infos = All information you want to know is in keys below
  • infos/author = Markus Raab elektra@markus-raab.org
  • infos/licence = BSD
  • infos/provides = resolver
  • infos/needs =
  • infos/placements = rollback getresolver setresolver commit
  • infos/status = productive maintained specific unittest tested libc nodep configurable
  • infos/description = system independent resolver


The @PLUGIN_SHORT_NAME@ handles operating system dependent tasks. One task is the resolving of the filenames for user and system (hence its name).

Use following command to see to which file is resolved to:

kdb file <Elektra path you are interested in>

See the constants of this plugin for further information, they are:


The build-in resolving considers following cases:

  • for spec with absolute path: path
  • for spec with relative path: KDB_DB_SPEC + path
  • for dir with absolute path: pwd + path (or above when path is found)
  • for dir with relative path: pwd + KDB_DB_DIR + path (or above when path is found)
  • for user with absolute path: ~ + path
  • for user with relative path: ~ + KDB_DB_USER + path
  • for system with absolute path: path
  • for system with relative path: KDB_DB_SYSTEM + path

(~ and pwd are resolved from system)


For an absolute path /example.ini, you might get following values:

  • for spec: /example.ini
  • for dir: pwd/example.ini
  • for user: ~/example.ini
  • for system: /example.ini

For an relative path example.ini, you might get following values:

  • for spec: /usr/share/elektra/specification/example.ini
  • for dir: pwd/.dir/example.ini
  • for user: ~/.config/example.ini
  • for system: /etc/kdb/example.ini

See the mount tutorial for more examples.


Many variants exist that additionally influence the resolving process, typically by using environment variables.

Environment variables are very simple for one-time usage but their maintenance in start-up scripts is problematic. Additionally, they are controlled by the user, so they cannot be trusted. So it is not recommended to use environment variables.

Note that the file permissions apply, so it might be possible for non-root to modify keys in system.

See COMPILE.md for a documentation of possible variants.

XDG Compatibility

The resolver is fully XDG compatible if configured with the variant:

  • xp, xh or xu for user (either using passwd, HOME or USER as fallback or any combination of these fallbacks)
  • x for system, no fallback necessary

Additionally KDB_DB_USER needs to be left unchanged as .config.

XDG_CONFIG_DIRS will be used to resolve system paths the following way:

  • if unset or empty /etc/xdg will be used instead
  • all elements are searched in order of importance
  • if a file was found, the search process is stopped
  • if no file was found, the least important element will be used for potential write attempts.

Reading Configuration

  1. If no update needed (unchanged modification time): quit successfully
  2. Otherwise call (storage) plugin(s) to read configuration
  3. remember the last stat time (last update)

Writing Configuration

  1. On unchanged configuration: quit successfully
  2. On empty configuration: remove the configuration file and quit successfully
  3. Otherwise, open the configuration file If not available recursively create directories and retry. #ifdef ELEKTRA_LOCK_MUTEX
  4. Try to lock a global mutex, if not possible -> conflict #endif #ifdef ELEKTRA_LOCK_FILE
  5. Try to lock the configuration file, if not possible -> conflict #endif
  6. Check the update time -> might lead to conflict
  7. Update the update time (in order to not self-conflict)

We have an optimistic approach. Locking is only used to detect concurrent cooperative processes in the short moment between prepare and commit. A conflict will be raised in that situation. When processes do not lock the file it might be overwritten. This is, however, very unlikely on file systems with nanosecond precision.

Exported Functions and Data

The resolver provides 2 functions for other plugins to use.


resolves path in with namespace namespace, creates a temporary file to work on and returns a struct containing the resolved data.

Signature: ElektraResolved * filename (elektraNamespace namespace, const char * path, ElektraResolveTempfile tempFile, Key * warningsKey)

ElektraResolved and ElektraResolveTempfile are both defined in shared.h.

ElektraResolved is a struct with the following members:

  • relPath: contains the path relative to the namespace.
  • dirname: contains the parent directory of the resolved file.
  • fullPath: contains the full path of the resolved file.
  • tmpFile: contains the full path of the created temporary file.

ElektraResolveTempfile dictates if and where a temporarey file should be created. Possible values:

  • ELEKTRA_RESOLVER_TEMPFILE_NONE: don't create a temporary file.
  • ELEKTRA_RESOLVER_TEMPFILE_SAMEDIR: create a temporary file in the same directory as the resolved file.
  • ELEKTRA_RESOLVER_TEMPFILE_TMPDIR: create a temporary file in /tmp.


frees the handle returned by filename.

Signature: void * freeHandle (ElektraResolved * handle)

where handle is the handle returned by filename.


If none of the resolving techniques work, the resolver will fail during kdbOpen. This happens, for example, with the default resolver (ELEKTRA_VARIANT_USER hpu) if neither: $HOME, $USER, nor any home directory in /etc/passwd is set.