- adding tests using very large passwords in auth
- fixes general http sending to treat h3 like h2, and
not like http1.1
- eliminate H2_HEADER max definitions and use the commmon
DYN_HTTP_REQUEST everywhere, different limits do not help
- fix http2 handling of requests denied by nghttp2 on send
to immediately report the refused stream
Closes#11509
Add a `libcurlu`/`libcurltool` static library that is compiled only for
unit tests. We use `EXCLUDE_FROM_ALL` to make sure that they're not
built by default, they're only built if unit tests are built.
These libraries allow us to compile every unit test with CMake.
Closes#11446
The updated Curl_http_req_make and Curl_http_req_make2 functions spawned
a parameter called m_len. The AmigaOS networking headers, derived from
NetBSD, contain "#define m_len m_hdr.mh_len" which clashes with
this. Since we do not actually use mbuf, force the include file to be
ignored, removing the clash.
Closes#11479
Rename `close` and `connect` in `struct Curl_cftype` for
consistency and to avoid clashes with macros of the same name
(the standard AmigaOS networking connect() function is implemented
via a macro).
Closes#11491
- a regression introduced by c9ec851211
where optimization of small POST bodies leads to a new code path
for such uploads that did not trigger the "done sending" event
- add triggering this event for early "upload_done" situations
Fixes#11485Closes#11487
Reported-by: Aleksander Mazur
- refs #11449 where a segfault is reported when IP Eyeballing did
not immediately connect but made several attempts
- The transfer initiating the eyeballing was initialized too early,
leadding to references to the filter instance that was then
replaced in the subsequent eyeball attempts. That led to a use
after free in the buffer handling for the transfer
- transfers are initiated now more lazy (like in the ngtcp2 filter),
when the stream is actually opened
- suppress reporting on quiche event errors for "other" transfers
than the current one to not fail a transfer due to faults in
another one.
- revert recent return value handling for quiche_h3_recv_body()
to not indicate an error but an EAGAIN situation. We wish quiche
would document what functions return.
Fixes#11449Closes#11469
Reported-by: ウさん
- refs #11449 where weirdness in quiche multi connection tranfers was
observed
- fixes lookup of transfer for a quiche event to take the connection
into account
- formerly, a transfer with the same stream_id, but on another connection
could be found
Closes#11462
- refs #11426 where spurious stalls on large POST requests
are reported
- the issue seems to involve the following
* first stream on connection adds up to 64KB of POST
data, which is the max default HTTP/2 stream window size
transfer is set to HOLD
* initial SETTINGS from server arrive, enlarging the stream
window. But no WINDOW_UPDATE is received.
* curl stalls
- the fix un-HOLDs a stream on receiving SETTINGS, not
relying on a WINDOW_UPDATE from lazy servers
Closes#11450
After upgrading to 8.1.2 from 7.84.0, I found that sockets were being
closed without calling the fclosesocket callback if a request was
cancelled after the associated socket was created, but before the socket
was connected. This lead to an imbalance of fopensocket & fclosesocket
callbacks, causing problems with a custom event loop integration using
the multi-API.
This was caused by cf_socket_close() calling sclose() directly instead
of calling socket_close() if the socket was not active. For regular TCP
client connections, the socket is activated by cf_socket_active(), which
is only called when the socket completes the connect.
As far as I can tell, this issue has existed since 7.88.0. That is,
since the code in question was introduced by:
commit 71b7e01610
Author: Stefan Eissing <stefan@eissing.org>
Date: Fri Dec 30 09:14:55 2022 +0100
lib: connect/h2/h3 refactor
Closes#11439
- TARGET_OS_OSX is not always defined on macOS
- this leads to missing symbol Curl_macos_init()
- TargetConditionals.h seems to define these only when
dynamic targets are enabled (somewhere?)
- this PR fixes that on my macOS 13.4.1
- I have no clue why CI builds worked without it
Follow-up to c7308592fbCloses#11417
https://github.com/curl/curl/pull/7121 introduced a macOS system call
to `SCDynamicStoreCopyProxies`, which is invoked every time an IP
address needs to be resolved.
However, this system call is not thread-safe, and macOS will kill the
process if the system call is run first in a fork. To make it possible
for the parent process to call this once and prevent the crash, only
invoke this system call in the global initialization routine.
In addition, this change is beneficial because it:
1. Avoids extra macOS system calls for every IP lookup.
2. Consolidates macOS-specific initialization in a separate file.
Fixes#11252Closes#11254
- refs #11389 where IDLE timeouts on upload are reported
- reword ngtcp2 expiry handling to apply to both send+recv
calls into the filter
- EAGAIN uploads similar to the recent changes in HTTP/2, e.g.
report success only when send data was ACKed.
- HOLD sending of EAGAINed uploads to avoid cpu busy loops
- rename internal function for consistency with HTTP/2
implementation
Fixes#11389Closes#11390
Make sure the user and password for the second request is taken from the
redirected-to URL.
Add test case 899 to verify.
Reported-by: James Lucas
Fixes#11410Closes#11412
- not quite to infinity
- rewrote the implementation of our internal HTTP/1.x request
parsing to work with very large lines using dynbufs.
- new default limit is `DYN_HTTP_REQUEST`, aka 1MB, which
is also the limit of curl's general HTTP request processing.
Fixes#11405Closes#11407
- fix HTTP/2 check to not declare a connection dead when
the read attempt results in EAGAIN
- add H2-PROXY alive check as for HTTP/2 that was missing
and is needed
- add attach/detach around Curl_conn_is_alive() and remove
these in filter methods
- add checks for number of connections used in some test_10
proxy tunneling tests
Closes#11368
- refs #11357, where it was reported that HTTP/1.1 downgrades
no longer works
- fixed with suggested change
- added test_05_03 and a new handler in the curltest module
to reproduce that downgrades work
Fixes#11357Closes#11362
Reported-by: Jay Satiro
- refs #11203 where hyper was reported as being slow
- fixes hyper_executor_poll to loop until it is out of
tasks as advised by @seanmonstar in https://github.com/hyperium/hyper/issues/3237
- added a fix in hyper io handling for detecting EAGAIN
- added some debug logs to see IO results
- pytest http/1.1 test cases pass
- pytest h2 test cases fail on connection reuse. HTTP/2
connection reuse does not seem to work. Hyper submits
a request on a reused connection, curl's IO works and
thereafter hyper declares `Hyper: [1] operation was canceled: connection closed`
on stderr without any error being logged before.
Fixes#11203
Reported-by: Gisle Vanem
Advised-by: Sean McArthur
Closes#11344
- fixes#11242 where 100% CPU on uploads was reported
- fixes possible stalls on last part of a request body when
that information could not be fully send on the connection
due to an EAGAIN
- applies the same EGAIN handling to HTTP/2 proxying
Reported-by: Sergey Alirzaev
Fixed#11242Closes#11342
Since the socket is not connected then the call fails. When the call
fails, failf() is called to write an error message that is then
surviving and is returned when the *real* error occurs later. The
earlier, incorrect, error therefore hides the actual error message.
This could be seen in stderr for test 1007
Test 1007 has now been extended to verify the stderr message.
Closes#11332
Previously it would count the size of the entire outgoing request and
not just the size of only the Cookie: header field - which was the
intention.
This could make the check be off by several hundred bytes in some cases.
Closes#11331
- Use CURL_FORMAT_CURL_OFF_T where %zd was erroneously used for some
curl_off_t variables.
- Use %zu where %zd was erroneously used for some size_t variables.
Prior to this change some of the Windows CI tests were failing because
in Windows 32-bit targets have a 32-bit size_t and a 64-bit curl_off_t.
When %zd was used for some curl_off_t variables then only the lower
32-bits was read and the upper 32-bits would be read for part or all of
the next specifier.
Fixes https://github.com/curl/curl/issues/11327
Closes https://github.com/curl/curl/pull/11321
- Implement AUTH=+LOGIN for CURLOPT_LOGIN_OPTIONS to prefer plaintext
LOGIN over SASL auth.
Prior to this change there was no method to be able to fall back to
LOGIN if an IMAP server advertises SASL capabilities. However, this may
be desirable for e.g. a misconfigured server.
Per: https://www.ietf.org/rfc/rfc5092.html#section-3.2
";AUTH=<enc-auth-type>" looks to be the correct way to specify what
authenication method to use, regardless of SASL or not.
Closes https://github.com/curl/curl/pull/10041
`max_recv_speed` is `curl_off_t`, so using `size_t` might result in
-Wconversion GCC warnings for 32-bit `size_t`. Visible in the NetBSD
ARM autobuilds.
Closes https://github.com/curl/curl/pull/11312
- add an `id` long to Curl_easy, -1 on init
- once added to a multi (or its own multi), it gets
a non-negative number assigned by the connection cache
- `id` is unique among all transfers using the same
cache until reaching LONG_MAX where it will wrap
around. So, not unique eternally.
- CURLINFO_CONN_ID returns the connection id attached to
data or, if none present, data->state.lastconnect_id
- variables and type declared in tool for write out
Closes#11185
- ngtcp2 v0.16.0 asserts that timestamps passed to its function
will only ever increase.
- Use a context shared between ingress/egress operations that
uses a shared timestamp, regularly updated during calls.
Closes#11288
