Documentation/technical/api-error-handling.txt

   1 Error reporting in git
   2 ======================
   3
   4 `BUG`, `die`, `usage`, `error`, and `warning` report errors of
   5 various kinds.
   6
   7 - `BUG` is for failed internal assertions that should never happen,
   8   i.e. a bug in git itself.
   9
  10 - `die` is for fatal application errors.  It prints a message to
  11   the user and exits with status 128.
  12
  13 - `usage` is for errors in command line usage.  After printing its
  14   message, it exits with status 129.  (See also `usage_with_options`
  15   in the link:api-parse-options.html[parse-options API].)
  16
  17 - `error` is for non-fatal library errors.  It prints a message
  18   to the user and returns -1 for convenience in signaling the error
  19   to the caller.
  20
  21 - `warning` is for reporting situations that probably should not
  22   occur but which the user (and Git) can continue to work around
  23   without running into too many problems.  Like `error`, it
  24   returns -1 after reporting the situation to the caller.
  25
  26 These reports will be logged via the trace2 facility. See the "error"
  27 event in link:api-trace2.txt[trace2 API].
  28
  29 Customizable error handlers
  30 ---------------------------
  31
  32 The default behavior of `die` and `error` is to write a message to
  33 stderr and then exit or return as appropriate.  This behavior can be
  34 overridden using `set_die_routine` and `set_error_routine`.  For
  35 example, "git daemon" uses set_die_routine to write the reason `die`
  36 was called to syslog before exiting.
  37
  38 Library errors
  39 --------------
  40
  41 Functions return a negative integer on error.  Details beyond that
  42 vary from function to function:
  43
  44 - Some functions return -1 for all errors.  Others return a more
  45   specific value depending on how the caller might want to react
  46   to the error.
  47
  48 - Some functions report the error to stderr with `error`,
  49   while others leave that for the caller to do.
  50
  51 - errno is not meaningful on return from most functions (except
  52   for thin wrappers for system calls).
  53
  54 Check the function's API documentation to be sure.
  55
  56 Caller-handled errors
  57 ---------------------
  58
  59 An increasing number of functions take a parameter 'struct strbuf *err'.
  60 On error, such functions append a message about what went wrong to the
  61 'err' strbuf.  The message is meant to be complete enough to be passed
  62 to `die` or `error` as-is.  For example:
  63
  64         if (ref_transaction_commit(transaction, &err))
  65                 die("%s", err.buf);
  66
  67 The 'err' parameter will be untouched if no error occurred, so multiple
  68 function calls can be chained:
  69
  70         t = ref_transaction_begin(&err);
  71         if (!t ||
  72             ref_transaction_update(t, "HEAD", ..., &err) ||
  73             ret_transaction_commit(t, &err))
  74                 die("%s", err.buf);
  75
  76 The 'err' parameter must be a pointer to a valid strbuf.  To silence
  77 a message, pass a strbuf that is explicitly ignored:
  78
  79         if (thing_that_can_fail_in_an_ignorable_way(..., &err))
  80                 /* This failure is okay. */
  81                 strbuf_reset(&err);