api-string-list.txt: initialize the string_list the easy way
[gitweb.git] / commit.c
index 8248a994a50ab91874600567e544aa3735e4aa98..87268682f98f0cd0f18221b3647ad30dbcbc560e 100644 (file)
--- a/commit.c
+++ b/commit.c
@@ -68,7 +68,7 @@ struct commit *lookup_commit_reference_by_name(const char *name)
        unsigned char sha1[20];
        struct commit *commit;
 
-       if (get_sha1(name, sha1))
+       if (get_sha1_committish(name, sha1))
                return NULL;
        commit = lookup_commit_reference(sha1);
        if (!commit || parse_commit(commit))
@@ -1112,8 +1112,92 @@ int commit_tree(const struct strbuf *msg, unsigned char *tree,
        return result;
 }
 
+static int find_invalid_utf8(const char *buf, int len)
+{
+       int offset = 0;
+
+       while (len) {
+               unsigned char c = *buf++;
+               int bytes, bad_offset;
+
+               len--;
+               offset++;
+
+               /* Simple US-ASCII? No worries. */
+               if (c < 0x80)
+                       continue;
+
+               bad_offset = offset-1;
+
+               /*
+                * Count how many more high bits set: that's how
+                * many more bytes this sequence should have.
+                */
+               bytes = 0;
+               while (c & 0x40) {
+                       c <<= 1;
+                       bytes++;
+               }
+
+               /* Must be between 1 and 5 more bytes */
+               if (bytes < 1 || bytes > 5)
+                       return bad_offset;
+
+               /* Do we *have* that many bytes? */
+               if (len < bytes)
+                       return bad_offset;
+
+               offset += bytes;
+               len -= bytes;
+
+               /* And verify that they are good continuation bytes */
+               do {
+                       if ((*buf++ & 0xc0) != 0x80)
+                               return bad_offset;
+               } while (--bytes);
+
+               /* We could/should check the value and length here too */
+       }
+       return -1;
+}
+
+/*
+ * This verifies that the buffer is in proper utf8 format.
+ *
+ * If it isn't, it assumes any non-utf8 characters are Latin1,
+ * and does the conversion.
+ *
+ * Fixme: we should probably also disallow overlong forms and
+ * invalid characters. But we don't do that currently.
+ */
+static int verify_utf8(struct strbuf *buf)
+{
+       int ok = 1;
+       long pos = 0;
+
+       for (;;) {
+               int bad;
+               unsigned char c;
+               unsigned char replace[2];
+
+               bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
+               if (bad < 0)
+                       return ok;
+               pos += bad;
+               ok = 0;
+               c = buf->buf[pos];
+               strbuf_remove(buf, pos, 1);
+
+               /* We know 'c' must be in the range 128-255 */
+               replace[0] = 0xc0 + (c >> 6);
+               replace[1] = 0x80 + (c & 0x3f);
+               strbuf_insert(buf, pos, replace, 2);
+               pos += 2;
+       }
+}
+
 static const char commit_utf8_warn[] =
-"Warning: commit message does not conform to UTF-8.\n"
+"Warning: commit message did not conform to UTF-8.\n"
 "You may want to amend it after fixing the message, or set the config\n"
 "variable i18n.commitencoding to the encoding your project uses.\n";
 
@@ -1170,7 +1254,7 @@ int commit_tree_extended(const struct strbuf *msg, unsigned char *tree,
        strbuf_addbuf(&buffer, msg);
 
        /* And check the encoding */
-       if (encoding_is_utf8 && !is_utf8(buffer.buf))
+       if (encoding_is_utf8 && !verify_utf8(&buffer))
                fprintf(stderr, commit_utf8_warn);
 
        if (sign_commit && do_sign_commit(&buffer, sign_commit))