Map SSL errors to SSL_PROTOCOL_ERROR, others to FAILED.

Map SSL errors to SSL_PROTOCOL_ERROR, others to FAILED.

This matches the NSS port's behavior. If an unknown error within ERR_LIB_SSL,
map to ERR_SSL_PROTOCOL_ERROR. Otherwise, map to ERR_FAILED. The OpenSSL port
actually inverted this for some reason.

(This doesn't fix #417134 but maps it to a more appropriate error.)

BUG=417134
Committed: https://crrev.com/5e114758a150ac5b7c0e025dc6a2392080b46e5e
Cr-Commit-Position: refs/heads/master@{#296769}

[davidben, 2014-09-24 15:23:00]

davidben@chromium.org changed reviewers:
+ agl@chromium.org

[agl, 2014-09-24 16:53:33]

LGTM.

SSL_PROTOCOL_ERROR is a magic value for fallbacks (see
http_network_transaction.cc). We have a test for fallback behaviour I believe
so, as long as that still passes, LGTM. None the less, this probably alters the
conditions that we'll do fallback in.

[davidben, 2014-09-25 19:33:16]

On 2014/09/24 16:53:33, agl wrote:
> LGTM.
> 
> SSL_PROTOCOL_ERROR is a magic value for fallbacks (see
> http_network_transaction.cc). We have a test for fallback behaviour I believe
> so, as long as that still passes, LGTM. None the less, this probably alters
the
> conditions that we'll do fallback in.

(Yeah, I actually extended some of those tests. :-) We used to be relying on
some of the ERR_SSL_PROTOCOL_FAILEDs which are now ERR_FAILEDs with this CL for
some of our fallback. But now that transport errors are preserved, it's fine and
the tests still pass.)

This does change the fallback cases somewhat, including when the bug in question
is tickled. Though apparently the bug also breaks fallback itself, so it doesn't
save us. Anyway, this should be more consistent with the NSS behavior; NSS also
falls back on any generic SSL protocol error.

[davidben, 2014-09-25 19:33:36]

The CQ bit was checked by davidben@chromium.org

[commit-bot: I haz the power, 2014-09-25 19:35:15]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/598043002/1

[commit-bot: I haz the power, 2014-09-25 19:41:09]

Committed patchset #1 (id:1) as 9c8c2c122fa1f163385e4bfb4c556d2efdabe6ba

[commit-bot: I haz the power, 2014-09-25 19:42:17]

Patchset 1 (id:??) landed as
https://crrev.com/5e114758a150ac5b7c0e025dc6a2392080b46e5e
Cr-Commit-Position: refs/heads/master@{#296769}