| Age | Commit message (Collapse) | Author |
|---|
|
All 1xx (informational), 204 (no content), and 304 (not modified)
responses MUST NOT include a message-body, and thus are always
terminated by the first empty line after the header fields.
This implies that chunked encoding MUST NOT be used for these
status codes.
This commit updates the client to gracefully handle responses from
faulty server implementations that can send chunked encoded 204,
304 or 1xx responses.
Change-Id: I2dd502e28b3c6e121640083118fa5c3e479f5194
|
|
Change-Id: I44fe2670e36884e09600d17dd71b1e86b9ee75fa
|
|
- Set gen_server call timeout to infinity in httpc_handler.
- Add new testcase 'slow_connection'.
Change-Id: Id11b54c588e1d66b9ebba4da5dbfe5e9bee1f1ee
|
|
Only run test case for http as the socket manipulation done by the test case
requires much more manipulation to work for https. However, that this test case
started failing constantly instead of sporadically is proof that this change
is a good thing.
|
|
Guard test for error case was not updated to handle httpc stream concept
properly.
|
|
|
|
|
|
Probably since 6153ba7 (OTP R13B04) the httpc timeout setting does not
work for redirects (when autoredirect is true). With this patch a new
timer is started for the new (redirected) requests. This means that a
simple redirected request could return after 2*timeout milliseconds.
This is the first part to fix https://bugs.erlang.org/browse/ERL-420
|
|
Input should be checked and httpc_handler process shall terminate
gracefully on errors so that user process will in hang in gen_server:call.
|
|
Remove dead code and redundant debug macros. Better to use tracing for
debugging.
|
|
Correct chunk decoding by adding missing argument to match.
The symptom was that chunk decoding sometimes failed depending on
stream data arrival timing.
|
|
|
|
|
|
* ingela/inets/httpc/ERL-253:
inets: httpc improve error handling
|
|
This supersedes PR #1185 (submitted by @KrzysiekJ) that changed
all occurrences of ref() with reference() in inets files.
However, there is little point in having these types only in comments.
So, these types are now exposed as type declarations for the record
fields they appear. While at it, uncommented more commented out type
declarations and declared types for records defined in the affected
modules and header files.
Some type-unfriendly and obsolete code related to supporting code
ungrades with a really old OTP release was also removed.
|
|
Handler process should exit normal if manager process has been terminated
|
|
httpc should fill SNI extenstion for HTTPS requests sent through
CONNECT tunnel to provide proper access to websites using SNI
|
|
Without this fix, httpc:request/1 crashes the httpc_handler when an
invalid chunked length header is encountered (since 77acb47):
=ERROR REPORT==== 14-Nov-2015::17:19:30 ===
** Generic server <0.651.0> terminating
** Last message in was {tcp,#Port<0.5714>,
<<"HTTP/1.1 200 ok\r\nTransfer-Encoding:chunked\r\n\r\nåäö\r\n">>}
** When Server state == {state,
[...]
** Reason for termination ==
** {bad_return_value,{error,{chunk_size,"åäö"}}}
|
|
httpc should work properly if streaming option {self, once} is
chosen and the corresponding response does not get streamed.
|
|
httpc_handler should respond with correct and complete
responses seeing non-streamed status codes i.e. codes
other than 200 or 206.
|
|
httpc should return meaningful error reason disregarding whether
it is being used with http proxy or not
|
|
httpc_handler should react properly to cancel requests even
when the request to be cancelled was already finished
but httpc_manager did not get notified about that yet
|
|
|
|
If the socket is closed by the peer do not try to close it again.
Solves OTP-11845
|
|
httpc_handler should set correct address of the connected host
and not ignore proxy option of the profile while
sending requests taken out of keep-alive queue
|
|
httpc_handler should respond with proper {stop, Reason, State}
message when sending request from keep-alive queue fails
|
|
* erlang/pr/741:
Minor misprints
|
|
|
|
The new module is backwards compatible.
|
|
httpc_handler should be stopped when sending requests returns error
because of a closed connection and `tcp_closed` message could not be
catched (see http://erlang.org/pipermail/erlang-bugs/2007-May/000346.html )
otherwise it will lead to process leak.
|
|
* scrapinghub/update_session_original_stacktrace:
inets: Return correct stacktrace on session updating failure
|
|
Debug output in httpc_handler:update_session/4 error
handling code should not pollute stack trace returned in process exit.
|
|
|
|
|
|
* scrapinghub/stream_body_with_no_content_length:
inets: Fix HTTP 1.0 body end on closed connection
|
|
* scrapinghub/stream_for_chunked_single_message:
inets: Fix streaming with single chunk body
|
|
Receiving HTTP response with no Content-length header and
with body ending on closed connection should give the complete
response body contents up to the last byte received.
|
|
Receiving HTTP response with chunked transfer encoding in a single
TCP message should produce stream messages for response body when
streaming mode is used for httpc.
|
|
* nox/httpc-transfer-encoding-identity:
Support identity transfer-encoding in httpc
OTP-11802
|
|
* ia/inets/revert/OTP-11756:
inets: Continue CTify test suite
inets: Enable https tests
inets: Make test suites independent of each other
inets: Reverted commmit f97cd6965ac35d1cbe510de23592956857d144c8
|
|
* scrapinghub/update_session_error_logging_fix:
inets: Fix incorrect argument order
OTP-11794
|
|
The reverted commit took a too optimistic approch to pipeling. When added
we thought it would be an improvment, but it proved to be a
misconception.
|
|
> The default (identity) encoding; the use of no transformation whatsoever.
> The Internet Assigned Numbers Authority (IANA) acts as a registry for
> transfer-coding value tokens. Initially, the registry contains the
> following tokens: "chunked" (section 3.6.1), "identity" (section 3.6.2),
> "gzip" (section 3.5), "compress" (section 3.5), and "deflate" (section 3.5).
Reported-by: Andrew Houghton
|
|
Wrong order of arguments in httpc_handler:update_session/4 error
handling code should not result in bad_argument error making
stack trace unusable.
|
|
The types array(), dict(), digraph(), gb_set(), gb_tree(), queue(),
set(), and tid() have been deprecated. They will be removed in OTP 18.0.
Instead the types array:array(), dict:dict(), digraph:graph(),
gb_set:set(), gb_tree:tree(), queue:queue(), sets:set(), and ets:tid()
can be used. (Note: it has always been necessary to use ets:tid().)
It is allowed in OTP 17.0 to locally re-define the types array(), dict(),
and so on.
New types array:array/1, dict:dict/2, gb_sets:set/1, gb_trees:tree/2,
queue:queue/1, and sets:set/1 have been added.
|
|
|
|
|
|
|
|
http_chunk:decode/4, the side-effecting chunk decoder, was only used
by httpc_handler:handle_http_body/2 but now it's completely unused.
|
|
Previously the only difference between {stream, self} and {stream,
{self, once}} was an extra Pid in the stream_start message due to a
bug in httpc_handler. It was immediately sending a bunch of messages
till the end instead of waiting for httpc:stream_next/1 being called.
|
