Reject unadvertised encodings

net/base/net_error_list.h

Index: net/base/net_error_list.h
diff --git a/net/base/net_error_list.h b/net/base/net_error_list.h
index e7a9052586d1da07689f53f7f1a067916f21a720..804c25db57371529f5b2455b1503e2c49716c8c2 100644
--- a/net/base/net_error_list.h
+++ b/net/base/net_error_list.h
@@ -702,6 +702,9 @@ NET_ERROR(CONTENT_DECODING_INIT_FAILED, -371)
 // SpdyStream layer.
 NET_ERROR(SPDY_RST_STREAM_NO_ERROR_RECEIVED, -372)
 
+// Content decoding of the response body failed.
+NET_ERROR(INVALID_CONTENT_ENCODING, -373)

[Randy Smith (Not in Mondays), 2017/03/15 16:35:31]

I"m inclined instead to make this CONTENT_DECODING_FAILURE and update the
Net.ContentDecodingFailed2.FilterType histogram to include "unknown" (may need
to be at a different site than the current one).

[eustas, 2017/03/16 12:08:16]

Done. Yup, net/http -> net/filter dependency doesn't look natural...

[Randy Smith (Not in Mondays), 2017/03/16 17:50:43]

I've gone back and forth on whether to ask you to do this or not, but: Ryan S &
I have basically agreed that a net/http -> net/filter dependency is a fine one,
and I'd rather have the histogram recordings just come from a single place, so:
Could you create a static function for recording decoding failures on
FilterSourceStream and call it from that function and from here to unify the
locations that this histogram is recorded from?  I do think the dependency is ok
(good, even), and I'd rather have it clear in code terms that there are multiple
places that are adding to this histogram.

Sorry about the extra work :-J.

[eustas, 2017/03/20 13:05:18]

No worries =)

+
 // The cache does not have the requested entry.
 NET_ERROR(CACHE_MISS, -400)