Christopher Head [Wed, 25 Jul 2018 06:40:11 +0000 (23:40 -0700)]
docs/CURLOPT_URL: fix indentation
The statement, “The application does not have to keep the string around
after setting this option,” appears to be indented under the RTMP
paragraph. It actually applies to all protocols, not just RTMP.
Eliminate the extra indentation.
Christopher Head [Wed, 25 Jul 2018 05:47:23 +0000 (22:47 -0700)]
docs/CURLOPT_WRITEFUNCTION: size is always 1
For compatibility with `fwrite`, the `CURLOPT_WRITEFUNCTION` callback is
passed two `size_t` parameters which, when multiplied, designate the
number of bytes of data passed in. In practice, CURL always sets the
first parameter (`size`) to 1.
This practice is also enshrined in documentation and cannot be changed
in future. The documentation states that the default callback is
`fwrite`, which means `fwrite` must be a suitable function for this
purpose. However, the documentation also states that the callback must
return the number of *bytes* it successfully handled, whereas ISO C
`fwrite` returns the number of items (each of size `size`) which it
wrote. The only way these numbers can be equal is if `size` is 1.
Since `size` is 1 and can never be changed in future anyway, document
that fact explicitly and let users rely on it.
wolfSSL/CyaSSL: Fix memory leak in Curl_cyassl_random
RNG structure must be freed by call to FreeRng after its use in
Curl_cyassl_random. This call fixes Valgrind failures when running the
test suite with wolfSSL.
Even Rouault [Wed, 25 Jul 2018 09:22:51 +0000 (11:22 +0200)]
reuse_conn(): free old_conn->options
This fixes a memory leak when CURLOPT_LOGIN_OPTIONS is used, together with
connection reuse.
I found this with oss-fuzz on GDAL and curl master:
https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=9582
I couldn't reproduce with the oss-fuzz original test case, but looking
at curl source code pointed to this well reproducable leak.
auth: pick Bearer authentication whenever a token is available
So far, the code tries to pick an authentication method only if
user/password credentials are available, which is not the case for
Bearer authentictation...
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Closes #2754
auth: only ever pick CURLAUTH_BEARER if we *have* a Bearer token
The Bearer authentication was added to cURL 7.61.0, but there is a
problem: if CURLAUTH_ANY is selected, and the server supports multiple
authentication methods including the Bearer method, we strongly prefer
that latter method (only CURLAUTH_NEGOTIATE beats it), and if the Bearer
authentication fails, we will never even try to attempt any other
method.
This is particularly unfortunate when we already know that we do not
have any Bearer token to work with.
Such a scenario happens e.g. when using Git to push to Visual Studio
Team Services (which supports Basic and Bearer authentication among
other methods) and specifying the Personal Access Token directly in the
URL (this aproach is frequently taken by automated builds).
Let's make sure that we have a Bearer token to work with before we
select the Bearer authentication among the available authentication
methods.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Closes #2754
Daniel Stenberg [Fri, 20 Jul 2018 21:24:47 +0000 (23:24 +0200)]
vtls: set conn->data when closing TLS
Follow-up to 1b76c38904f0. The VTLS backends that close down the TLS
layer for a connection still needs a Curl_easy handle for the session_id
cache etc.
Daniel Stenberg [Thu, 19 Jul 2018 22:57:28 +0000 (00:57 +0200)]
smb_getsock: always wait for write socket too
... the protocol is doing read/write a lot, so it needs to write often
even when downloading. A more proper fix could check for eactly when it
wants to write and only ask for it then.
Without this fix, an SMB download could easily get stuck when the event-driven
API was used.
By default, the MSYS2 bash interprets http:/%HOSTIP:%HTTPPORT/want/1143
as a POSIX file list and converts it to a Windows file list.
Disable this with MSYS2_ARG_CONV_EXCL for the test to pass.
Patrick Monnerat [Thu, 12 Jul 2018 20:46:15 +0000 (22:46 +0200)]
content_encoding: accept up to 4 unknown trailer bytes after raw deflate data
Some servers issue raw deflate data that may be followed by an undocumented
trailer. This commit makes curl tolerate such a trailer of up to 4 bytes
before considering the data is in error.
The definition of CALG_TLS1PRF has been fixed in the 5.1 branch:
https://osdn.net/projects/mingw/scm/git/mingw-org-wsl/commits/73aedcc0f2e6ba370de0d86ab878ad76a0dda7b5
Daniel Stenberg [Wed, 4 Jul 2018 12:47:10 +0000 (14:47 +0200)]
multi: always do the COMPLETED procedure/state
It was previously erroneously skipped in some situations.
libtest/libntlmconnect.c wrongly depended on wrong behavior (that it
would get a zero timeout) when no handles are "running" in a multi
handle. That behavior is no longer present with this fix. Now libcurl
will always return a -1 timeout when all handles are completed.
Paul Howarth [Wed, 11 Jul 2018 11:01:38 +0000 (12:01 +0100)]
openssl: assume engine support in 1.0.0 or later
Commit 38203f1585da changed engine detection to be version-based,
with a baseline of openssl 1.0.1. This does in fact break builds
with openssl 1.0.0, which has engine support - the configure script
detects that ENGINE_cleanup() is available - but <openssl/engine.h>
doesn't get included to declare it.
According to upstream documentation, engine support was added to
mainstream openssl builds as of version 0.9.7:
https://github.com/openssl/openssl/blob/master/README.ENGINE
This commit drops the version test down to 1.0.0 as version 1.0.0d
is the oldest version I have to test with.
Original MinGW's w32api has a sytax error in its definition of
CALG_TLS1PRF [0]. Don't use original MinGW w32api's CALG_TLS1PRF
until this bug [1] is fixed.
When size_t is not a typedef for unsigned long (as usually the case on
Windows), GCC emits -Wformat warnings when using lu and lx format
specifiers with size_t. Silence them with explicit casts to
unsigned long.
Daniel Stenberg [Thu, 28 Jun 2018 21:24:21 +0000 (23:24 +0200)]
openssl: make the requested TLS version the *minimum* wanted
The code treated the set version as the *exact* version to require in
the TLS handshake, which is not what other TLS backends do and probably
not what most people expect either.
Reported-by: Andreas Olsson Assisted-by: Gaurav Malhotra
Fixes #2691
Closes #2694
Adrian Peniak [Thu, 28 Jun 2018 05:56:16 +0000 (06:56 +0100)]
CURLINFO_TLS_SSL_PTR.3: improve the example
The previous example was a little bit confusing, because SSL* structure
(or other "in use" SSL connection pointer) is not accessible after the
transfer is completed, therefore working with the raw TLS library
specific pointer needs to be done during transfer.
Daniel Stenberg [Tue, 26 Jun 2018 16:11:32 +0000 (18:11 +0200)]
easy_perform: use *multi_timeout() to get wait times
... and trim the threaded Curl_resolver_getsock() to return zero
millisecond wait times during the first three milliseconds so that
localhost or names in the OS resolver cache gets detected and used
faster.
Max Dymond [Tue, 26 Jun 2018 08:29:48 +0000 (09:29 +0100)]
configure: Add dependent libraries after crypto
The linker is pretty dumb and processes things left to right, keeping a
tally of symbols it hasn't resolved yet. So, we need -ldl to appear
after -lcrypto otherwise the linker won't find the dl functions.
Daniel Stenberg [Wed, 20 Jun 2018 21:00:36 +0000 (23:00 +0200)]
url: fix dangling conn->data pointer
By masking sure to use the *current* easy handle with extracted
connections from the cache, and make sure to NULLify the ->data pointer
when the connection is put into the cache to make this mistake easier to
detect in the future.
Daniel Stenberg [Thu, 14 Jun 2018 14:28:59 +0000 (16:28 +0200)]
multi: fix memory leak when stopped during name resolve
When the application just started the transfer and then stops it while
the name resolve in the background thread hasn't completed, we need to
wait for the resolve to complete and then cleanup data accordingly.
Enabled test 1553 again and added test 1590 to also check when the host
name resolves successfully.
Mamta Upadhyay [Thu, 14 Jun 2018 16:29:43 +0000 (09:29 -0700)]
maketgz: fix sed issues on OSX
maketgz creates release tarballs and removes the -DEV string in curl
version (e.g. 7.58.0-DEV), else -DEV shows up on command line when curl
is run. maketgz works fine on linux but fails on OSX. Problem is with
the sed commands that use option -i without an extension. Maketgz
expects GNU sed instead of BSD and this simply won't work on OSX. Adding
a backup extension .bak after -i fixes this issue
projects / curl / log