Re: [PATCH v7 02/40] builtin/apply: make apply_patch() return -1 instead of die()ing

On Tue, Jun 14, 2016 at 12:55 AM, Junio C Hamano [off-list ref] wrote:

Christian Couder [off-list ref] writes:

quoted

+/*
+ * Try to apply a patch.
+ *
+ * Returns:
+ *  -1 if an error happened
+ *   0 if the patch applied
+ *   1 if the patch did not apply
+ */
 static int apply_patch(struct apply_state *state,
                     int fd,
                     const char *filename,

@@ -4413,6 +4421,7 @@ static int apply_patch(struct apply_state *state,
      struct strbuf buf = STRBUF_INIT; /* owns the patch text */
      struct patch *list = NULL, **listp = &list;
      int skipped_patch = 0;
+     int res = 0;

      state->patch_input_file = filename;
      read_patch_file(&buf, fd);

@@ -4445,8 +4454,10 @@ static int apply_patch(struct apply_state *state,
              offset += nr;
      }

-     if (!list && !skipped_patch)
-             die(_("unrecognized input"));
+     if (!list && !skipped_patch) {
+             res = error(_("unrecognized input"));
+             goto end;
+     }

Before this patch, the program said "fatal: $message" and exited
with status = 128.  All these changes in this step modifies the
external behaviour and make it say "error: $message" and exit with
status = 1 (at least the caller in apply_all_patches() does so).

Will that be an issue for the calling scripts?

Hopefully the scripts don't check the specific error code and message.

I will add something about this in the commit message.

Do you think something else that should be done about this?