On Fri, Dec 29, 2017 at 08:45:57PM +0000, Lucas Rolff wrote:
> So, this Wx1ZSI usually should be "cookie" - however it's somehow turned into
> garbage.
Ah, this is what I was wondering.
> Repeated - now it's s3U2JV - but still supposed to be "cookie":
Great, so the number of characters is correct but not the decoding. I'll
instrument the huffman decoder to dump the original code. It's possible
that we have a bug there. It passed all the 26k tests I found, but it
doesn't mean it never breaks! Another possibility would be that this
one is passed as "never index" and that we'd have a bug there.
Here comes a patch that will dump the raw hpack frame so that I can run
it by hand and spot of we're getting something wrong. Beware, it *will*
reveal every raw header your browser sends including the domain name
or authentication tokens if any. If you don't feel comfortable with
sending this to the list, please just send this to me in private. It
should be compatible with the current patch, helping to sort it out.
Thanks a lot for your help!
Willy
diff --git a/src/hpack-dec.c b/src/hpack-dec.c
index 454f55c..d5d282c 100644
--- a/src/hpack-dec.c
+++ b/src/hpack-dec.c
@@ -155,6 +155,8 @@ int hpack_decode_frame(struct hpack_dht *dht, const uint8_t
*raw, uint32_t len,
int must_index;
int ret;
+ debug_hexdump(stderr, "[HPACK-DEC] ", raw, 0, len);
+
chunk_reset(tmp);
ret = 0;
while (len) {
