[Bug 37844] Don't try to reconnect an account over and over for certain errors

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=37844

Emilio Pozuelo Monfort <pochu27 at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #48028|0                           |1
        is obsolete|                            |

--- Comment #8 from Emilio Pozuelo Monfort <pochu27 at gmail.com> 2011-06-16 04:40:35 PDT ---
Created an attachment (id=48037)
 View: https://bugs.freedesktop.org/attachment.cgi?id=48037
 Review: https://bugs.freedesktop.org/review?bug=37844&attachment=48037

Check for the specific error codes when reconnecting

(In reply to comment #7)
> Review of attachment 48028 [details]:
> 
> ::: src/mcd-connection.c
> @@ +1217,3 @@
> +            /* possibly a few others (notably, all the encryption
> +             * things), but the others are hopefully not
> +             * generically mapped to NETWORK_ERROR anyway */
> 
> Er, that comment was meant to be a note to you, not something you should commit
> :-)

Oh :-)

> Either replace it with more error codes (look at the list in devhelp or
> something and think about whether each one would be helped by reconnection or
> not), or check that for all our other error codes, believing the status reason
> would be OK.

I have added quite a few. There are a bunch that I don't think would ever
happen when connecting (e.g. errors that are channel-specific), so I have left
those out.

I have added another DEBUG in case we don't handle a TpError.

> @@ +1241,3 @@
> +    {
> +    case TP_CONNECTION_STATUS_REASON_NETWORK_ERROR:
> +    case TP_CONNECTION_STATUS_REASON_NONE_SPECIFIED:
> 
> I'm not so sure that NONE_SPECIFIED should trigger reconnection: if the string
> error name was not understood and the status reason is NONE_SPECIFIED, that
> tells us nothing.
> 
> The comment suggests that NONE_SPECIFIED was only there because it's what a
> crashing connection manager would produce, but TP_DBUS_ERROR_NAME_OWNER_LOST
> covers that case now, so maybe we don't need it?

Agreed, I've removed it. If we find out that there are other cases that it was
covering, we should add the specific TpErrors if possible.

> @@ +1245,3 @@
> +        return TRUE;
> +    default:
> +        break;
> 
> Useless default case that does nothing, can hopefully be removed?

I added it because gcc was complaining.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
You are the assignee for the bug.