On 02/12/2013 03:14 PM, Michal Toman wrote:
+    safe_waitpid(xz_child, &status, 0);
+    if (!WIFEXITED(status) || WEXITSTATUS(status) != 0)
+    {
+        error_msg_and_die(_("Can't create temporary file in /tmp"));
+    }
        ^^^ I can see some inconsistency here - on many places you put
the curly braces around a single function and on many you don't.

+ err:
+    alert_server_error(NULL);
+    /* Show bad header to the user */
+    char *sanitized = sanitize_utf8(message, (SANITIZE_ALL &
~SANITIZE_TAB));
Do we really need to sanitize? IIRC HTTP headers are never UTF-8. Either
they are valid (ASCII) or totally broken and should probably not be
displayed.

-void alert_server_error();
-void alert_connection_error();
+void alert_server_error(const char *peer_name);
+void alert_connection_error(const char *peer_name);
You use 'peer_name' in the header and 'name' in the implementation.

On 11.02.2013 19:39, Denys Vlasenko  wrote:
Open questions:

(1)
      alert_connection_error(URL);
      error_msg_and_die("Something horrible happened at '%s'", URL);

looks redundant. It looks like we _want_ error_msg_and_die
to also perform the duty of alert_connection_error.
alert_* stuff is meant for non-technical GUI (GNOME) users, who do not
care about the actual error. error_msg_and_die prints the useful message
to log. I don't insist on this approach if we can find a better solution
(I actually agree that it's weird).

(2) alert() needs to be made variadic a-la printf.
Didn't seem necessary so far, but may be extended.

(3) I have a nagging desire to nuke all trailing periods in messages...
Whatever, I have no strong opinion.

- don't forgot to update the translation if you touch some strings
$ tx push -s

--J



M.

Reply via email to