Update IP related information at the connection and the transfer in two
places only: once the filter chain connects and when a transfer is added
to a connection. The latter only updates on reuse when the filters
already are connected.
The only user of that information before a full connect is the HAProxy
filter. Add cfilter CF_QUERY_IP_INFO query to let it find the
information from the filters "below".
This solves two issues with the previous version:
- updates where often done twice with the same info
- happy eyeballing filter "forks" could overwrite each others
updates before the full winner was determined.
Closes#14699
When we downloaded all we wanted, and we did not want a response body,
and no Trailer: has been announced, and the receive gives EAGAIN, do not
hang around unnecessarily.
Some servers are buggy in HEAD processing and fail to send the HTTP/2
EOS. Since we do not need any more data, end the request right there.
This will cause us to send a RST_STREAM to the server.
Fixes#14670
Reported-by: Gruber Glass
Closes#14685
This is a better match for what they do and the general "cpool"
var/function prefix works well.
The pool now handles very long hostnames correctly.
The following changes have been made:
* 'struct connectdata', e.g. connections, keep new members
named `destination` and ' destination_len' that fully specifies
interface+port+hostname of where the connection is going to.
This is used in the pool for "bundling" of connections with
the same destination. There is no limit on the length any more.
* Locking: all locks are done inside conncache.c when calling
into the pool and released on return. This eliminates hazards
of the callers keeping track.
* 'struct connectbundle' is now internal to the pool. It is no
longer referenced by a connection.
* 'bundle->multiuse' no longer exists. HTTP/2 and 3 and TLS filters
no longer need to set it. Instead, the multi checks on leaving
MSTATE_CONNECT or MSTATE_CONNECTING if the connection is now
multiplexed and new, e.g. not conn->bits.reuse. In that case
the processing of pending handles is triggered.
* The pool's init is provided with a callback to invoke on all
connections being discarded. This allows the cleanups in
`Curl_disconnect` to run, wherever it is decided to retire
a connection.
* Several pool operations can now be fully done with one call.
Pruning dead connections, upkeep and checks on pool limits
can now directly discard connections and need no longer return
those to the caller for doing that (as we have now the callback
described above).
* Finding a connection for reuse is now done via `Curl_cpool_find()`
and the caller provides callbacks to evaluate the connection
candidates.
* The 'Curl_cpool_check_limits()' now directly uses the max values
that may be set in the transfer's multi. No need to pass them
around. Curl_multi_max_host_connections() and
Curl_multi_max_total_connections() are gone.
* Add method 'Curl_node_llist()' to get the llist a node is in.
Used in cpool to verify connection are indeed in the list (or
not in any list) as they need to.
I left the conncache.[ch] as is for now and also did not touch the
documentation. If we update that outside the feature window, we can
do this in a separate PR.
Multi-thread safety is not achieved by this PR, but since more details
on how pools operate are now "internal" it is a better starting
point to go for this in the future.
Closes#14662
1. GHA/windows: enable WinIDN in Linux cross-builds.
(to reveal the issue in CI.)
2. fix compiler warning when building with mingw-w64 supporting
WinIDN, while targeting pre-Vista Windows, with a `WINVER` set to
target Vista or newer. (Such was Ubuntu's mingw-w64 with the
classic-mingw-specific trick in point 3 of this PR.)
```
../../lib/idn.c:154:23: error: redundant redeclaration of ‘IdnToAscii’ [-Werror=redundant-decls]
154 | WINBASEAPI int WINAPI IdnToAscii(DWORD dwFlags,
| ^~~~~~~~~~
In file included from /usr/share/mingw-w64/include/windows.h:73,
from /usr/share/mingw-w64/include/winsock2.h:23,
from ../../lib/setup-win32.h:91,
from ../../lib/curl_setup.h:308,
from ../../lib/idn.c:29:
/usr/share/mingw-w64/include/winnls.h:1075:30: note: previous declaration of ‘IdnToAscii’ was here
1075 | WINNORMALIZEAPI int WINAPI IdnToAscii (DWORD dwFlags, LPCWSTR lpUnicodeCharStr, int cchUnicodeChar, LPWSTR lpASCIICharStr, int cchASCIIChar);
| ^~~~~~~~~~
[...same for IdnToUnicode...]
```
Ref: https://github.com/curl/curl/actions/runs/10542832783/job/29210098553#step:7:89
3. drop `WINVER` override for classic-mingw. curl no longer supports
building with classic-mingw.
Reverts 37f1c21cb9#7581
4. sync `if IdnToUnicode can be linked` detection snippet with the live
code in `lib/idn.c`. It fixes detection for the scenario in point 2.
5. delete unused `WINIDN_DIR` variable.
Bug: https://github.com/curl/curl/pull/12606#issuecomment-1885381038
Previous abandoned attempt: #12684
Reviewed-by: Jay Satiro
Closes#14680
Before this patch, `libhostname.so` and `chkhostname` were a test
facility for overriding `gethostname()` in non-debug builds on
Linux and other Unix platforms supporting `LD_PRELOAD`.
`gethostname()` has a single use with SMTP.
The alternative way to override `gethostname()` is building in debug
mode, which allows to do this via the `CURL_GETHOSTNAME` env, on all
platforms.
Drop the `LD_PRELOAD` solution in favour of the above.
Also:
- delete inactive NTLM code with a `gethostname()` call made from it.
- streamline NTLM code by dropping a `printf()` and a macro.
- tests: stop setting `CURL_GETHOSTNAME` where unnecessary.
Closes#14695
Before this change, calling curl_share_setopt w/ CURL_LOCK_DATA_CONNECT
a second time would re-initialize the connection cache, rather than use
the existing one.
After this change, calling curl_share_setopt w/ CURL_LOCK_DATA_CONNECT
multiple times will have no effect after the first call.
Closes#14696
Curl_xfer_send and Curl_xfer_recv had commented FIXMEs about protocol
setting up the transfers badly, but in reality these functions are too
low-level to be able to depend on the protocol transfer setups having
been done yet. Removed.
The functions had checks for data and data->conn that I convered to
asserts since they SHOULD always be valid in this function. The same
goes for the runtime check for buffer_size > 0 that I also converted to
an assert since that should never be set to an invalid value.
Closes#14688
Remove the "hardcoded" logic for the pop3 transfer handler and instead
use the generic protocol handler write_resp function.
Remove the check for 'data->req.ignorebody' because I cannot find a code
flow where this is set for POP3.
Closes#14684
- configure: disable pthreads by default on Windows.
- configure: disable detecting `fseeko()` on Windows.
(It exists in mingw-w64 2.0.0 and newer, but it's permanently ignored
in CMake, as this function is never necessary on Windows.)
- extend existing exceptions with their Windows variants.
- `lib/formdata.c`: prioritize `_fseeki64()` over `fseeko()`.
To reduce the difference between Windows builds, which now all use
`_fseeki64()`.
- cmake: perm-enable `HAVE_DIRENT_H` and `HAVE_OPENDIR` for mingw-w64,
to match configure.
Follow-up to bfe54b0e88#13137
This in theory could make the dir listing feature work in mingw-w64
build, but in my tests (on WINE) it failed at the preceding `open()`
call.
- cmake: perm-enable `HAVE_STRINGS_H` and `HAVE_UTIME_H` for mingw-w64,
to match configure. (They are wrappers and make no difference in the build.)
Also:
- configure: sync `USE_MANUAL` macro with cmake, by only setting it for
`src`. Drop checker exception.
- CI: use `--disable-dependency-tracking` in existing jobs.
- CI: install packages before git checkout, in existing jobs.
Closes#14678
- show `OpenSSL v3+` when detected (as in `./configure`).
(this string also makes its way to `curl-config`.)
- prefer `unset(VAR)` over `set(VAR)`.
Same effect, but `unset()` tells the intent unambiguously.
https://cmake.org/cmake/help/latest/command/set.html
- drop "implementation" from an `option()` description.
- FindGSS: replace legacy keyword alias with modern alternative.
https://cmake.org/cmake/help/latest/command/get_filename_component.html
- move `CURL_STATIC_CRT` logic next to its `option()`.
- improve order of `libcurl.pc`/`curl-config` variable init lines.
- tests: drop/shorten custom target names.
They inflated generated make files by 550KB.
Keep target name logic for sync between code snippets.
Follow-up to a2ef5d36b3#14660
- clear a variable after use.
- restore `STATUS` for `Features:`/`Protocols:` `message()`s:
Without it the output goes to stderr, and appears in red in CMake GUI.
It doesn't seem possible to show a line on stdout without leading
underscores to match `curl -V` and `./configure` output.
Partial revert of acbc6b703f#14197
- WindowsCache: move `HAVE_LINUX_TCP_H` into the header group.
- move strings to the same line as their `STRING` keyword.
- formatting in generated code.
- delete bogus comment.
- unfold lines for readability.
- fix a too long line. (for cmakelint)
- missing quotes, whitespace, comments.
Closes#14610
It was previously wrongly verifying the input in its URL encoded format
when setting the hostname component with curl_url_set(), so it wrongly
rejected '%'.
Now it URL decodes the name appropriately before the check.
Added tests to lib1560 to verify that a fine %-code is okay and that a
bad %-code (that decodes to '%') is rejected.
Regression from 0a0c9b6dfa, shipped in 8.0.0
Fixes#14656
Reported-by: Venkat Krishna R
Closes#14657
Previously this test allowed several error values when setting options.
This made this test miss #14629.
Now, errors are generally not accepted for setopts:
- numerical setopts accept CURLE_BAD_FUNCTION_ARGUMENT for funny input
- the first setopt to an option accepts CURLE_NOT_BUILT_IN or
CURLE_UNKNOWN_OPTION for when they are disabled/not built-in
- there is an allowlist concept for some return code for some variables,
managed at the top of the mk-lib1521.pl script
In curl.h: remove the OBSOLETE named values from the setopt list.
Closes#14634
Sources used `lib/curlx.h` with both `ENABLE_CURLX_PRINTF` set and unset
before including it.
In a cmake "unity" batch where the first included source had it unset,
the next sources did not get the macros requested with
`ENABLE_CURLX_PRINTF` because `lib/curl.x` had already been included
without them.
Fix it by by making the macros enabled permanently and globally for
internal sources, and dropping `ENABLE_CURLX_PRINTF`.
This came up while testing unity builds with smaller batches. The full,
default unity build where all `src` is bundled up in a single unit, was
not affected.
Fixes:
```
$ cmake -B build -DCMAKE_UNITY_BUILD=ON -DCMAKE_UNITY_BUILD_BATCH_SIZE=15
$ make -C build
...
curl/src/tool_getparam.c: In function ‘getparameter’:
curl/src/tool_getparam.c:2409:11: error: implicit declaration of function ‘msnprintf’; did you mean ‘vsnprintf’? [-Wimplicit-function-declaration]
2409 | msnprintf(buffer, sizeof(buffer), "%" CURL_FORMAT_CURL_OFF_T "-",
| ^~~~~~~~~
| vsnprintf
curl/src/tool_getparam.c:2409:11: warning: nested extern declaration of ‘msnprintf’ [-Wnested-externs]
[...]
```
Reported-by: Daniel Stenberg
Bug: https://github.com/curl/curl/pull/14626#issuecomment-2301663491Closes#14632
```
/Users/runner/work/curl/curl/lib/version.c: In function 'curl_version_info':
/Users/runner/work/curl/curl/lib/version.c:584:15: error: declaration of 'ssh_buffer' shadows a global declaration [-Werror=shadow]
584 | static char ssh_buffer[80];
| ^~~~~~~~~~
In file included from /Users/runner/work/curl/curl/lib/vssh/ssh.h:35,
from /Users/runner/work/curl/curl/lib/urldata.h:185,
from /Users/runner/work/curl/curl/lib/altsvc.c:32,
from /Users/runner/work/curl/curl/bld/lib/CMakeFiles/libcurl_shared.dir/Unity/unity_0_c.c:4:
/opt/homebrew/include/libssh/libssh.h:99:35: note: shadowed declaration is here
99 | typedef struct ssh_buffer_struct* ssh_buffer;
| ^~~~~~~~~~
```
Ref: https://github.com/curl/curl/actions/runs/10477958747/job/29020250670#step:9:48
Tested via #14614Closes#14617
For libssh, it fixes a "unity" build issue where libssh deprecation
warnings were not suppressed before this patch, because the suppression
macro was only set before just one of the two `libssh.h` includes.
If the other was compiled first in unity mode, the warnings appeared.
Seen in local curl-for-win build (`CW_CONFIG=test-x64-libssh-quictls`)
with libssh 0.11.0. (Also in a GHA/macos cmake job upcoming in #14614)
Use this opportunity to drop duplicate SSH header includes from the SSH
modules. It's enough to include them via the common `ssh.h` header.
Closes#14612
With mbedTLS if the minimum version of TLS is set to 1.3,
MBEDTLS_SSL_VERIFY_OPTIONAL is not available in client mode. See:
https://github.com/Mbed-TLS/mbedtls/blob/2ca6c285/library/ssl_tls.c#L1357
Also, there might be plans to remove it completely in future mbedTLS
versions.
Switch to always use MBEDTLS_SSL_VERIFY_REQUIRED. If verifypeer or
verifyhost are disabled the corresponding error flags are cleared in the
verify callback function. That is also where verification errors are
logged.
Closes#14591
As the data might be held by TLS buffers, leaving some and expecting to
get called again is error prone.
Reported-by: ralfjunker on github
Fixes#14201Closes#14597
After TLS handshare, indicate which TLS version was negotiated in
addition to the cipher in the handshake completed log message.
Also use the verify callback for certificate logging and collection.
This allows things to work even when MBEDTLS_SSL_KEEP_PEER_CERTIFICATE
is disabled in the mbedtls library.
And lastly, catch certificate validation errors later so we can give the
user more informative error messages that indicate what the failure was
from certificate validation.
Tested on both current LTS versions (2.28 and 3.6).
Closes#14444
- Renames Curl_readwrite() to Curl_sendrecv() to reflect that it
is mainly about talking to the server, not reads or writes to the
client. Add a `nowp` parameter since the single caller already
has this.
- Curl_sendrecv() now runs all possible operations whenever it is
called and either it had been polling sockets or the 'select_bits'
are set.
POLL_IN/POLL_OUT are not always directly related to send/recv
operations. Filters like HTTP/2, QUIC or TLS may monitor reverse
directions. If a transfer does not want to send (KEEP_SEND), it
will not do so, as before. Same for receives.
- Curl_update_timer() now checks the absolute timestamp of an expiry
and the last/new timeout to determine if the application needs
to stop/start/restart its timer. This fixes edge cases where
updates did not happen as they should have.
- improved --test-event curl_easy_perform() simulation to handle
situations where no sockets are registered but a timeout is
in place.
- fixed bug in events_socket() that complained about removing
a socket that was unknown, when indeed it had removed the socket
just before, only it was the last in the list
- fixed conncache's internal handle to carry the multi instance
(where the cache has one) so that operations on the closure handle
trigger event callbacks correctly.
- fixed conncache to not POLL_REMOVE a socket twice when a conneciton
was closed.
Closes#14561
- move variable dump to a GHA foldable group.
- minimize scope for an include().
- rename `HIDES_CURL_PRIVATE_SYMBOLS` to `CURL_HIDES_PRIVATE_SYMBOLS`,
to keep it in the curl namespace.
- drop quotes from a version number.
- add missing `Makefile.inc` var refs to comment.
- FindNGTCP2: rename internal var to underscore/lowercase.
- FindBearSSL, FindGSS: whitespace.
Closes#14571
Enable with CMake option `-DCURL_USE_WOLFSSH=ON`. Customize with
`-DWOLFSSH_INCLUDE_DIR=<path-to-wolfssh>/include`,
`-DWOLFSSH_LIBRARY=<path-to-wolfssh>/lib/libwolfssh.a`.
It requires the wolfSSL TLS backend.
Closes#14568
- add set/get functions for the custom data in a tree node
- use Curl_timediff for time comparisons instead of the custom macro, as they
do the exact same things.
- add asserts to catch mistakes better
- updated test 1309 accordingly
Closes#14562
Previously an error from the callback accidentally made libcurl keep the
hash entry which would lead to the entry remaining and then potentially
getting removed *again* which would lead to internal confusions.
This is an old issue (introduced in 2b3dd01b77), caught by the new
asserts from c0233a35da.
Closes#14557
Curl_mime_read() may go into an infinite loop when called with buffer
lengths <= 4. Some encoders, like base64, are not prepared for that.
In the client reader for mime data, skip such small reads. The upload
buffers will get flushed eventually and larger reads will happen again.
Improves robustness of test652 which triggered the loop on blocked
sends.
Closes#14532
