Daniel Stenberg [Tue, 22 Nov 2022 07:25:50 +0000 (08:25 +0100)]
lib: rewind BEFORE request instead of AFTER previous
This makes a big difference for cases when the rewind is not actually
necessary to perofm (for example HTTP response code 301 converts to GET)
and therefore the rewind can be avoided. In particular for situations
when that rewind fails, for example when reading from a pipe or similar.
Reported-by: Ali Utku Selen
Fixes #9735
Closes #9958
openssl: silence compiler warning when not using IPv6
In non-IPv6 builds the conn parameter is unused, and compilers which
run with "-Werror=unused-parameter" (or similar) warnings turned on
fails to build. Below is an excerpt from a CI job:
Stefan Eissing [Tue, 22 Nov 2022 08:55:41 +0000 (09:55 +0100)]
vtls: localization of state data in filters
- almost all backend calls pass the Curl_cfilter intance instead of
connectdata+sockindex
- ssl_connect_data is remove from struct connectdata and made internal
to vtls
- ssl_connect_data is allocated in the added filter, kept at cf->ctx
- added function to let a ssl filter access its ssl_primary_config and
ssl_config_data this selects the propert subfields in conn and data,
for filters added as plain or proxy
- adjusted all backends to use the changed api
- adjusted all backends to access config data via the exposed
functions, no longer using conn or data directly
cfilter renames for clear purpose:
- methods `Curl_conn_*(data, conn, sockindex)` work on the complete
filter chain at `sockindex` and connection `conn`.
- methods `Curl_cf_*(cf, ...)` work on a specific Curl_cfilter
instance.
- methods `Curl_conn_cf()` work on/with filter instances at a
connection.
- rebased and resolved some naming conflicts
- hostname validation (und session lookup) on SECONDARY use the same
name as on FIRST (again).
new debug macros and removing connectdata from function signatures where not
needed.
adapting schannel for new Curl_read_plain paramter.
Daniel Stenberg [Tue, 22 Nov 2022 08:32:42 +0000 (03:32 -0500)]
examples/10-at-a-time: fix possible skipped final transfers
Prior to this change if curl_multi_perform returned 0 running handles
and then all remaining transfers were added, then the perform loop would
end immediately without performing those transfers.
Viktor Szakats [Tue, 22 Nov 2022 08:25:05 +0000 (08:25 +0000)]
Makefile.mk: portable Makefile.m32
Update bare GNU Make `Makefile.m32` to:
- Move objects into a subdirectory.
- Add support for MS-DOS. Tested with DJGPP.
- Add support for Watt-32 (on MS-DOS).
- Add support for AmigaOS.
- Rename `Makefile.m32` to `Makefile.mk`
- Replace `ARCH` with `TRIPLET`.
- Build `tool_hugehelp.c` proper (when tools are available).
- Drop MS-DOS compatibility macro `USE_ZLIB` (replaced by `HAVE_LIBZ`)
- Add support for `ZLIB_LIBS` to override `-lz`.
- Omit object files when building examples.
- Default `CC` to `gcc` once again, for convenience. (Caveat: compiler
name `cc` cannot be set now.)
- Set `-DCURL_NO_OLDIES` for examples, like autotools does.
- Delete `makefile.dj` files. Notice the configuration details and
defaults are not retained with the new method.
- Delete `makefile.amiga` files. A successful build needs a few custom
options. We're also not retaining all build details from the existing
Amiga make files.
- Rename `Makefile.m32` to `Makefile.mk` to reflect that they are not
Windows/MinGW32-specific anymore.
- Add support for new `CFG` options: `-map`, `-debug`, `-trackmem`
- Set `-DNDEBUG` by default.
- Allow using `-DOS=...` in all `lib/config-*.h` headers, syncing this
with `config-win32.h`.
- Look for zlib parts in `ZLIB_PATH/include` and `ZLIB_PATH/lib`
instead of bare `ZLIB_PATH`.
Note that existing build configurations for MS-DOS and AmigaOS likely
become incompatible with this change.
Example AmigaOS configuration:
```
export CROSSPREFIX=/opt/amiga/bin/m68k-amigaos-
export CC=gcc
export CPPFLAGS='-DHAVE_PROTO_BSDSOCKET_H'
export CFLAGS='-mcrt=clib2'
export LDFLAGS="${CFLAGS}"
export LIBS='-lnet -lm'
make -C lib -f Makefile.mk
make -C src -f Makefile.mk
```
Example MS-DOS configuration:
```
export CROSSPREFIX=/opt/djgpp/bin/i586-pc-msdosdjgpp-
export WATT_PATH=/opt/djgpp/net/watt
export ZLIB_PATH=/opt/djgpp
export OPENSSL_PATH=/opt/djgpp
export OPENSSL_LIBS='-lssl -lcrypt'
export CFG=-zlib-ssl
make -C lib -f Makefile.mk
make -C src -f Makefile.mk
```
Jay Satiro [Mon, 14 Nov 2022 08:30:30 +0000 (03:30 -0500)]
sendf: change Curl_read_plain to wrap Curl_recv_plain (take 2)
Prior to this change Curl_read_plain would attempt to read the
socket directly. On Windows that's a problem because recv data may be
cached by libcurl and that data is only drained using Curl_recv_plain.
Rather than rewrite Curl_read_plain to handle cached recv data, I
changed it to wrap Curl_recv_plain, in much the same way that
Curl_write_plain already wraps Curl_send_plain.
This fixes a bug in the schannel backend where decryption of arbitrary
TLS records fails because cached recv data is never drained. We send
data (TLS records formed by Schannel) using Curl_write_plain, which
calls Curl_send_plain, and that may do a recv-before-send
("pre-receive") to cache received data. The code calls Curl_read_plain
to read data (TLS records from the server), which prior to this change
did not call Curl_recv_plain and therefore cached recv data wasn't
retrieved, resulting in malformed TLS records and decryption failure
(SEC_E_DECRYPT_FAILURE).
The bug has only been observed during Schannel TLS 1.3 handshakes. Refer
to the issue and PR for more information.
--
This is take 2 of the original fix. It preserves the original behavior
of Curl_read_plain to write 0 to the bytes read parameter on error,
since apparently some callers expect that (SOCKS tests were hanging).
The original fix which landed in 12e1def5 and was later reverted in 18383fbf failed to work properly because it did not do that.
Also, it changes Curl_write_plain the same way to complement
Curl_read_plain, and it changes Curl_send_plain to return -1 instead of
0 on CURLE_AGAIN to complement Curl_recv_plain.
Behavior on error with these changes:
Curl_recv_plain returns -1 and *code receives error code.
Curl_send_plain returns -1 and *code receives error code.
Curl_read_plain returns error code and *n (bytes read) receives 0.
Curl_write_plain returns error code and *written receives 0.
HTTP/3 needs a special filter chain, since it does the TLS handling
itself. This PR adds special setup handling in the HTTP protocol handler
that takes are of it.
When a handler, in its setup method, installs filters, the default
behaviour for managing the filter chain is overridden.
Reported-by: Karthikdasari0423 on github
Fixes #9931
Closes #9945
Jay Satiro [Mon, 14 Nov 2022 08:30:30 +0000 (03:30 -0500)]
sendf: change Curl_read_plain to wrap Curl_recv_plain
Prior to this change Curl_read_plain would attempt to read the
socket directly. On Windows that's a problem because recv data may be
cached by libcurl and that data is only drained using Curl_recv_plain.
Rather than rewrite Curl_read_plain to handle cached recv data, I
changed it to wrap Curl_recv_plain, in much the same way that
Curl_write_plain already wraps Curl_send_plain.
This fixes a bug in the schannel backend where decryption of arbitrary
TLS records fails because cached recv data is never drained. We send
data (TLS records formed by Schannel) using Curl_write_plain, which
calls Curl_send_plain, and that may do a recv-before-send
("pre-receive") to cache received data. The code calls Curl_read_plain
to read data (TLS records from the server), which prior to this change
did not call Curl_recv_plain and therefore cached recv data wasn't
retrieved, resulting in malformed TLS records and decryption failure
(SEC_E_DECRYPT_FAILURE).
The bug has only been observed during Schannel TLS 1.3 handshakes. Refer
to the issue and PR for more information.
Jay Satiro [Mon, 14 Nov 2022 09:07:30 +0000 (04:07 -0500)]
test3026: reduce runtime in legacy mingw builds
- Load Windows system libraries secur32 and iphlpapi beforehand, so
that libcurl's repeated global init/cleanup only increases/decreases
the library's refcount rather than causing it to load/unload.
Assisted-by: Marc Hoersken
Closes https://github.com/curl/curl/pull/9412
Daniel Stenberg [Thu, 17 Nov 2022 22:55:26 +0000 (23:55 +0100)]
url: move back the IDN conversion of proxy names
Regression: in commit 53bcf55 we moved the IDN conversion calls to
happen before the HSTS checks. But the HSTS checks are only done on the
server host name, not the proxy names. By moving the proxy name IDN
conversions, we accidentally broke the verbose output showing the proxy
name.
This change moves back the IDN conversions for the proxy names to the
place in the code path they were before 53bcf55.
Patrick Monnerat [Mon, 14 Nov 2022 14:21:34 +0000 (15:21 +0100)]
lib: feature deprecation warnings in gcc >= 4.3
Add a deprecated attribute to functions and enum values that should not
be used anymore.
This uses a gcc 4.3 dialect, thus is only available for this version of
gcc and newer. Note that the _Pragma() keyword is introduced by C99, but
is available as part of the gcc dialect even when compiling in C89 mode.
It is still possible to disable deprecation at a calling module compile
time by defining CURL_DISABLE_DEPRECATION.
Gcc type checking macros are made aware of possible deprecations.
Some testing support Perl programs are adapted to the extended
declaration syntax.
Several test and unit test C programs intentionally use deprecated
functions/options and are annotated to not generate a warning.
New test 1222 checks the deprecation status in doc and header files.
Daniel Stenberg [Sun, 13 Nov 2022 22:58:47 +0000 (23:58 +0100)]
cmdline-opts/gen.pl: fix the linkifier
Improved logic for finding existing --options in text and replacing with
the full version with nroff syntax. This also makes the web version link
options better.
Reported-by: xianghongai on github
Fixes #9899
Closes #9902
tool: use feature names instead of bit mask, when possible
If the run-time libcurl is too old to support feature names, the name
array is created locally from the bit masks. This is the only sequence
left that uses feature bit masks.
Stefan Eissing [Fri, 11 Nov 2022 14:15:37 +0000 (15:15 +0100)]
Websocket: fixes for partial frames and buffer updates.
- buffers updated correctly when handling partial frames
- callbacks no longer invoked for incomplete payload data of 0 length
- curl_ws_recv no longer returns with 0 length partial payload
Viktor Szakats [Fri, 11 Nov 2022 20:12:15 +0000 (20:12 +0000)]
config-win32: fix SIZEOF_OFF_T for MSVC and old MinGW
The previously set default value of 8 (64-bit) is only correct for
mingw-w64 and only when we set `_FILE_OFFSET_BITS` to 64 (the default
when building curl). For MSVC, old MinGW and other Windows compilers,
the correct value is 4 (32-bit). Adjust condition accordingly. Also
drop the manual override option.
Stefan Eissing [Fri, 11 Nov 2022 10:45:34 +0000 (11:45 +0100)]
lib: connection filters (cfilter) addition to curl:
- general construct/destroy in connectdata
- default implementations of callback functions
- connect: cfilters for connect and accept
- socks: cfilter for socks proxying
- http_proxy: cfilter for http proxy tunneling
- vtls: cfilters for primary and proxy ssl
- change in general handling of data/conn
- Curl_cfilter_setup() sets up filter chain based on data settings,
if none are installed by the protocol handler setup
- Curl_cfilter_connect() boot straps filters into `connected` status,
used by handlers and multi to reach further stages
- Curl_cfilter_is_connected() to check if a conn is connected,
e.g. all filters have done their work
- Curl_cfilter_get_select_socks() gets the sockets and READ/WRITE
indicators for multi select to work
- Curl_cfilter_data_pending() asks filters if the have incoming
data pending for recv
- Curl_cfilter_recv()/Curl_cfilter_send are the general callbacks
installed in conn->recv/conn->send for io handling
- Curl_cfilter_attach_data()/Curl_cfilter_detach_data() inform filters
and addition/removal of a `data` from their connection
- adding vtl functions to prevent use of Curl_ssl globals directly
in other parts of the code.
Fata Nugraha [Wed, 9 Nov 2022 09:38:00 +0000 (16:38 +0700)]
http: do not send PROXY more than once
Unlike `CONNECT`, currently we don't keep track whether `PROXY` is
already sent or not. This causes `PROXY` header to be sent twice during
`MSTATE_TUNNELING` and `MSTATE_PROTOCONNECT`.
Jay Satiro [Tue, 8 Nov 2022 23:49:21 +0000 (18:49 -0500)]
lib: add CURL_WRITEFUNC_ERROR to signal write callback error
Prior to this change if the user wanted to signal an error from their
write callbacks they would have to use logic to return a value different
from the number of bytes (nmemb) passed to the callback. Also, the
inclination of some users has been to just return 0 to signal error,
which is incorrect as that may be the number of bytes passed to the
callback.
To remedy this the user can now return CURL_WRITEFUNC_ERROR instead.
Daniel Stenberg [Thu, 27 Oct 2022 08:21:27 +0000 (10:21 +0200)]
KNOWN_BUGS: remove eight entries
- 1.2 Multiple methods in a single WWW-Authenticate: header
This is not considered a bug anymore but a restriction and one that we
keep because we have NEVER gotten this reported by users in the wild and
because of this I consider this a fringe edge case we don't need to
support.
- 1.6 Unnecessary close when 401 received waiting for 100
This is not a bug, but possibly an optimization that *can* be done.
- 1.7 Deflate error after all content was received
This is not a curl bug. This happens due to broken servers.
- 2.1 CURLINFO_SSL_VERIFYRESULT has limited support
This is not a bug. This is just the nature of the implementation.
Daniel Stenberg [Fri, 4 Nov 2022 23:10:24 +0000 (00:10 +0100)]
curl: add --url-query
This option adds a piece of data, usually a name + value pair, to the
end of the URL query part. The syntax is identical to that used for
--data-urlencode with one extension:
If the argument starts with a '+' (plus), the rest of the string is
provided as-is unencoded.
This allows users to "build" query parts with options and URL encoding
even when not doing GET requests, which the already provided option -G
(--get) is limited to.
Viktor Szakats [Tue, 1 Nov 2022 22:45:28 +0000 (22:45 +0000)]
windows: fail early with a missing windres in autotools
`windres` is not always auto-detected by autotools when building for
Windows. When this happened, the build failed with a confusing error due
to the empty `RC` command:
Viktor Szakats [Tue, 1 Nov 2022 22:40:36 +0000 (22:40 +0000)]
lib: sync guard for Curl_getaddrinfo_ex() definition and use
`Curl_getaddrinfo_ex()` gets _defined_ with `HAVE_GETADDRINFO` set. But,
`hostip4.c` _used_ it with `HAVE_GETADDRINFO_THREADSAFE` set alone. It
meant a build with the latter, but without the former flag could result
in calling this function but not defining it, and failing to link.
Patch this by adding an extra check for `HAVE_GETATTRINFO` around the
call.
Before this patch, build systems prevented this condition. Now they
don't need to.
While here, simplify the related CMake logic on Windows by setting
`HAVE_GETADDRINFO_THREADSAFE` to the detection result of
`HAVE_GETADDRINFO`. This expresses the following intent clearer than
the previous patch and keeps the logic in a single block of code:
When we have `getaddrinfo()` on Windows, it's always threadsafe.
Viktor Szakats [Tue, 1 Nov 2022 22:27:28 +0000 (22:27 +0000)]
tidy-up: process.h detection and use
This patch aims to cleanup the use of `process.h` header and the macro
`HAVE_PROCESS_H` associated with it.
- `process.h` is always available on Windows. In curl, it is required
only for `_beginthreadex()` in `lib/curl_threads.c`.
- `process.h` is also available in MS-DOS. In curl, its only use was in
`lib/smb.c` for `getpid()`. But `getpid()` is in fact declared by
`unistd.h`, which is always enabled via `lib/config-dos.h`. So the
header is not necessary.
- `HAVE_PROCESS_H` was detected by CMake, forced to 1 on Windows and
left to real detection for other platforms.
It was also set to always-on in `lib/config-win32.h` and
`lib/config-dos.h`.
In autotools builds, there was no detection and the macro was never
set.
Based on these observations, in this patch we:
- Rework Windows `getpid` logic in `lib/smb.c` to always use the
equivalent direct Win32 API function `GetCurrentProcessId()`, as we
already did for Windows UWP apps. This makes `process.h` unnecessary
here on Windows.
- Stop #including `process.h` into files where it was not necessary.
This is everywhere, except `lib/curl_threads.c`.
> Strangely enough, `lib/curl_threads.c` compiled fine with autotools
> because `process.h` is also indirecty included via `unistd.h`. This
> might have been broken in autotools MSVC builds, where the latter
> header is missing.
- Delete all remaining `HAVE_PROCESS_H` feature guards, for they were
unnecessary.
- Delete `HAVE_PROCESS_H` detection from CMake and predefined values
from `lib/config-*.h` headers.
Patrick Monnerat [Mon, 31 Oct 2022 03:23:59 +0000 (04:23 +0100)]
gen.pl: do not generate CURLHELP bitmask lines > 79 characters
If a command line option is in many help categories, there is a risk
that CURLHELP bitmask source lines generated for listhelp are longer
than 79 characters.
This change takes care of folding such long lines.
projects / thirdparty / curl.git / log
