Re: [Qemu-devel] [PATCH 15/25] qemu-sockets: add error propagation to inet_connect_addr

[Markus Armbruster]

Paolo Bonzini <address@hidden> writes:

> Il 17/10/2012 17:40, Markus Armbruster ha scritto:
>>> >      if (s->current_addr) {
>>> >          while (s->current_addr->ai_next != NULL && s->fd < 0) {
>>> >              s->current_addr = s->current_addr->ai_next;
>>> > -            s->fd = inet_connect_addr(s->current_addr, &in_progress, s);
>>> > +            s->fd = inet_connect_addr(s->current_addr, &in_progress, s, 
>>> > NULL);
>> Doesn't this drop error messages?
>
> Depends on what you mean by drop.  The error messages previously were
> sent to stdio, hidden in a log file or just invisible, depending on how
> QEMU is run.  (This is just for outgoing migration currently, so you
> cannot rely on stdio as you can for command-line parsing).

I agree the existing error reporting is very bad, but even very bad
reporting is (slightly) better than no error reporting.

Moreover, there's a use case where errors are reported in the correct
place now: -monitor stdio.  I'm afraid your patch hurts it.

>> If yes, but it's healed later in the series, the temporary breakage
>> still needs to be spelled out in the commit message.
>
> No, it's not healed, which is why it's mentioned in the commit message
> that future work is needed.

I really hate "reward" submissions of tons of useful work with demands
for more even work...  but here goes anyway: what's your take on
completing the job?  Straightfoward, or another ton of work?