Standardize error reporting of IPEndPoint::ToSockAddr and

Standardize error reporting of IPEndPoint::ToSockAddr and
IPEndPoint::FromSockAddr.

R=eroman@chromium.org
BUG=247210
TEST=none

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=259719

[wtc, 2014-03-04 00:43:27]

Eric:

Before you review the CL, please help me resolve two issues.

1. IPEndPoint::ToSockAddr and IPEndPoint::FromSockAddr return bool, so if the
caller needs to return a network error code, the caller need to make up one.
Right now we return anything from ERR_FAILED, ERR_UNEXPECTED,
ERR_INVALID_ARGUMENT, to ERR_ADDRESS_INVALID. I think ERR_ADDRESS_INVALID is the
best error code because the only reasons IPEndPoint::ToSockAddr and
IPEndPoint::FromSockAddr can fail are unsupported address family and incorrect
address length for the address family. ERR_INVALID_ARGUMENT is also reasonable,
though less precise. (I found evidence that connect() and sendto() use EINVAL in
these two cases.)

2. If you agree that ERR_ADDRESS_INVALID is the best error code, I found that we
were originally using ERR_ADDRESS_INVALID for the OS error code EADDRNOTAVAIL,
which actually has a different meaning. I think we should map EADDRNOTAVAIL to a
new error code, ERR_ADDRESS_NOT_AVAILABLE.

https://codereview.chromium.org/185553013/diff/1/net/socket/tcp_socket_libeve...
File net/socket/tcp_socket_libevent.cc (left):

https://codereview.chromium.org/185553013/diff/1/net/socket/tcp_socket_libeve...
net/socket/tcp_socket_libevent.cc:828: errno = EINVAL;

Setting errno to EINVAL here was apparently an attempt to get MapSystemError()
to return ERR_INVALID_ARGUMENT later (see line 642 above). So if we want
MapSystemError() to return ERR_ADDRESS_INVALID instead, we'll need to set errno
to EADDRNOTAVAIL.

[eroman, 2014-03-04 03:01:09]

lgtm

Your suggestions sound good on both accounts.

https://codereview.chromium.org/185553013/diff/1/net/socket/tcp_socket_libeve...
File net/socket/tcp_socket_libevent.cc (left):

https://codereview.chromium.org/185553013/diff/1/net/socket/tcp_socket_libeve...
net/socket/tcp_socket_libevent.cc:828: errno = EINVAL;
On 2014/03/04 00:43:28, wtc wrote:
> 
> Setting errno to EINVAL here was apparently an attempt to get MapSystemError()
> to return ERR_INVALID_ARGUMENT later (see line 642 above). So if we want
> MapSystemError() to return ERR_ADDRESS_INVALID instead, we'll need to set
errno
> to EADDRNOTAVAIL.

That is kind of gross

[wtc, 2014-03-26 18:34:02]

The CQ bit was checked by wtc@chromium.org

[commit-bot: I haz the power, 2014-03-26 18:35:07]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/wtc@chromium.org/185553013/40001

[commit-bot: I haz the power, 2014-03-26 22:24:42]

Change committed as 259719