* HTTPS proxies:
An HTTPS proxy receives all transactions over an SSL/TLS connection.
Once a secure connection with the proxy is established, the user agent
uses the proxy as usual, including sending CONNECT requests to instruct
the proxy to establish a [usually secure] TCP tunnel with an origin
server. HTTPS proxies protect nearly all aspects of user-proxy
communications as opposed to HTTP proxies that receive all requests
(including CONNECT requests) in vulnerable clear text.
With HTTPS proxies, it is possible to have two concurrent _nested_
SSL/TLS sessions: the "outer" one between the user agent and the proxy
and the "inner" one between the user agent and the origin server
(through the proxy). This change adds supports for such nested sessions
as well.
A secure connection with a proxy requires its own set of the usual SSL
options (their actual descriptions differ and need polishing, see TODO):
--proxy-cacert FILE CA certificate to verify peer against
--proxy-capath DIR CA directory to verify peer against
--proxy-cert CERT[:PASSWD] Client certificate file and password
--proxy-cert-type TYPE Certificate file type (DER/PEM/ENG)
--proxy-ciphers LIST SSL ciphers to use
--proxy-crlfile FILE Get a CRL list in PEM format from the file
--proxy-insecure Allow connections to proxies with bad certs
--proxy-key KEY Private key file name
--proxy-key-type TYPE Private key file type (DER/PEM/ENG)
--proxy-pass PASS Pass phrase for the private key
--proxy-ssl-allow-beast Allow security flaw to improve interop
--proxy-sslv2 Use SSLv2
--proxy-sslv3 Use SSLv3
--proxy-tlsv1 Use TLSv1
--proxy-tlsuser USER TLS username
--proxy-tlspassword STRING TLS password
--proxy-tlsauthtype STRING TLS authentication type (default SRP)
All --proxy-foo options are independent from their --foo counterparts,
except --proxy-crlfile which defaults to --crlfile and --proxy-capath
which defaults to --capath.
Curl now also supports %{proxy_ssl_verify_result} --write-out variable,
similar to the existing %{ssl_verify_result} variable.
Supported backends: OpenSSL, GnuTLS, and NSS.
* A SOCKS proxy + HTTP/HTTPS proxy combination:
If both --socks* and --proxy options are given, Curl first connects to
the SOCKS proxy and then connects (through SOCKS) to the HTTP or HTTPS
proxy.
TODO: Update documentation for the new APIs and --proxy-* options.
Look for "Added in 7.XXX" marks.
As they are after all part of the public API. Saves space and reduces
complexity. Remove the strcase defines from the curlx_ family.
Suggested-by: Dan Fandrich
Idea: https://curl.haxx.se/mail/lib-2016-10/0136.html
... to make it less likely that we forget that the function actually
does case insentive compares. Also replaced several invokes of the
function with a plain strcmp when case sensitivity is not an issue (like
comparing with "-").
After a few wasted hours hunting down the reason for slowness during a
TLS handshake that turned out to be because of TCP_NODELAY not being
set, I think we have enough motivation to toggle the default for this
option. We now enable TCP_NODELAY by default and allow applications to
switch it off.
This also makes --tcp-nodelay unnecessary, but --no-tcp-nodelay can be
used to disable it.
Thanks-to: Tim Rühsen
Bug: https://curl.haxx.se/mail/lib-2016-06/0143.html
- Move the existing scheme check from tool_operate.
In the case of --remote-header-name we want to parse Content-disposition
for a filename, but only if the scheme is http or https. A recent
adjustment 0dc4d8e was made to account for schemeless URLs however it's
not 100% accurate. To remedy that I've moved the scheme check to the
header callback, since at that point the library has already determined
the scheme.
Bug: https://github.com/curl/curl/issues/760
Reported-by: Kai Noda
It does open up a miniscule risk that one of the other protocols that
libcurl could use would send back a Content-Disposition header and then
curl would act on it even if not HTTP.
A future mitigation for this risk would be to allow the callback to ask
libcurl which protocol is being used.
Verified with test 1312
Closes#760
The underlying libcurl option used for this feature is
CURLOPT_FTP_CREATE_MISSING_DIRS which has the ability to retry the dir
creation, but it was never set to do that by the command line tool.
Now it does.
Bug: https://curl.haxx.se/mail/archive-2016-04/0021.html
Reported-by: John Wanghui
Help-by: Leif W
As these two options provide identical functionality, the former for
SOCK5 proxies and the latter for HTTP proxies, merged the two options
together.
As such CURLOPT_SOCKS5_GSSAPI_SERVICE is marked as deprecated as of
7.49.0.
- Add tests.
- Add an example to CURLOPT_TFTP_NO_OPTIONS.3.
- Add --tftp-no-options to expose CURLOPT_TFTP_NO_OPTIONS.
Bug: https://github.com/curl/curl/issues/481
- Add unit test 1604 to test the sanitize_file_name function.
- Use -DCURL_STATICLIB when building libcurltool for unit testing.
- Better detection of reserved DOS device names.
- New flags to modify sanitize behavior:
SANITIZE_ALLOW_COLONS: Allow colons
SANITIZE_ALLOW_PATH: Allow path separators and colons
SANITIZE_ALLOW_RESERVED: Allow reserved device names
SANITIZE_ALLOW_TRUNCATE: Allow truncating a long filename
- Restore sanitization of banned characters from user-specified outfile.
Prior to this commit sanitization of a user-specified outfile was
temporarily disabled in 2b6dadc because there was no way to allow path
separators and colons through while replacing other banned characters.
Now in such a case we call the sanitize function with
SANITIZE_ALLOW_PATH which allows path separators and colons to pass
through.
Closes https://github.com/curl/curl/issues/624
Reported-by: Octavio Schroeder
Due to path separators being incorrectly sanitized in --output
pathnames, eg -o c:\foo => c__foo
This is a partial revert of 3017d8a until I write a proper fix. The
remote-name will continue to be sanitized, but if the user specified an
--output with string replacement (#1, #2, etc) that data is unsanitized
until I finish a fix.
Bug: https://github.com/bagder/curl/issues/624
Reported-by: Octavio Schroeder
curl does not sanitize colons in a remote file name that is used as the
local file name. This may lead to a vulnerability on systems where the
colon is a special path character. Currently Windows/DOS is the only OS
where this vulnerability applies.
CVE-2016-0754
Bug: http://curl.haxx.se/docs/adv_20160127B.html
Make this the default for the curl tool (if built with HTTP/2 powers
enabled) unless a specific HTTP version is requested on the command
line.
This should allow more users to get HTTP/2 powers without having to
change anything.
They didn't match the ifdef logic used within libcurl anyway so they
could indeed warn for the wrong case - plus the tool cannot know how the
lib actually performs at that level.
They tend to never get updated anyway so they're frequently inaccurate
and we never go back to revisit them anyway. We document issues to work
on properly in KNOWN_BUGS and TODO instead.
It uses 'Note:' as a prefix as opposed to the common 'Warning:' to take
down the tone a bit.
It adds a warning for using -XHEAD on other methods becasue that may
lead to a hanging connection.
It isn't always clear to the user which options that cause the HTTP
methods to conflict so by spelling them out it should hopefully be
easier to understand why curl complains.
- Add new option CURLOPT_DEFAULT_PROTOCOL to allow specifying a default
protocol for schemeless URLs.
- Add new tool option --proto-default to expose
CURLOPT_DEFAULT_PROTOCOL.
In the case of schemeless URLs libcurl will behave in this way:
When the option is used libcurl will use the supplied default.
When the option is not used, libcurl will follow its usual plan of
guessing from the hostname and falling back to 'http'.
New tool option --ssl-no-revoke.
New value CURLSSLOPT_NO_REVOKE for CURLOPT_SSL_OPTIONS.
Currently this option applies only to WinSSL where we have automatic
certificate revocation checking by default. According to the
ssl-compared chart there are other backends that have automatic checking
(NSS, wolfSSL and DarwinSSL) so we could possibly accommodate them at
some later point.
Bug: https://github.com/bagder/curl/issues/264
Reported-by: zenden2k <zenden2k@gmail.com>
Follow-up to e8423f9ce1 with discussionis in
https://github.com/bagder/curl/pull/258
This check scans for fopen() with a mode string without 'b' present, as
it may indicate that an FOPEN_* define should rather be used.
This commit fixes a regression introduced in curl-7_41_0-186-g261a0fe.
It also introduces a regression test 1424 based on tests 78 and 1423.
Reported-by: Viktor Szakats
Bug: https://github.com/bagder/curl/issues/237
As the 'error' and 'mute' options are now part of the GlobalConfig,
rather than per Operation, updated the warnf() function to use this
structure rather than the OperationConfig.
There was a mix of GlobCode, CURLcode and ints and they were mostly
passing around CURLcode errors. This change makes the functions use only
CURLcode and removes the GlobCode type completely.
Mark CURLOPT_UNIX_SOCKET_PATH as string to ensure that it ends up as
option in the file generated by --libcurl.
Signed-off-by: Peter Wu <peter@lekensteyn.nl>
Coverity CID 1243583. get_url_file_name() cannot fail and return a NULL
file name pointer so skip the check for that - it tricks coverity into
believing it can happen and it then warns later on when we use 'outfile'
without checking for NULL.
Option --pinnedpubkey takes a path to a public key in DER format and
only connect if it matches (currently only implemented with OpenSSL).
Provides CURLOPT_PINNEDPUBLICKEY for curl_easy_setopt().
Extract a public RSA key from a website like so:
openssl s_client -connect google.com:443 2>&1 < /dev/null | \
sed -n '/-----BEGIN/,/-----END/p' | openssl x509 -noout -pubkey \
| openssl rsa -pubin -outform DER > google.com.der
Ensure a source file isn't generated for the following informational
command line parameters when --libcurl is specified:
--help, --manual, --version and --engine list
As the output would only include a fairly empty looking main() function
and a call to curl_easy_init() and curl_easy_cleanup() when performed
with --engine list.
Correctly output libcurl source code that includes multiply operations
as specified by --next. Note that each operation evaluates to a single
curl_easy_perform() in source code form.
Also note that the output could be optimised a little so global config
options are only output once rather than per operation as is presently
the case.
In preparation for separating the global config options from the per
operation config options, reworked the list engines code to not use a
member variable in the Configurable structure.
To help assist with the detection of incorrect return codes, as per
commits ee23d13a79, 33b8960dc8 and aba98991a5, updated the operate
based functions to return CURLcode error codes.
During initialisation SetHTTPrequest() may fail and cURL would return
PARAM_BAD_USE, which is equivalent to CURLE_NOT_BUILT_IN in cURL error
terms.
Instead, return CURLE_FAILED_INIT as we do for other functions that may
fail during initialisation.
Rather than check for required arguments, and prompt for any host and
proxy passwords, as each operation is performed, changed the code so
all configurations are checked before any operations are performed.
This allows the user to input all the required passwords, for example,
upfront rather than wait for each operation.
Since protocol headers contain explicit line-endings there should
be no automatic conversion to ASCII text or CRLF line-endings.
This might break third party tools that already depend on this
behaviour. We might need to introduce an option to make this optional.
Commmit c5f8e2f5f4 removed the easy handle clean-up from tool_operate,
letting the code that was already present in free_config_fields()
perform the task. Unfortunately, this wasn't the correct place to do
this as it broke protocols, that would perform a logout, as the main
clean-up in tool_main had already been called.
when using --http2 one can now selectively disable NPN or ALPN with
--no-alpn and --no-npn. for now honored with NSS only.
TODO: honor this option with GnuTLS and OpenSSL
In the rare instance where getparameter() may return PARAM_NO_MEM whilst
parsing a URL, cURL would return this error code, which is equivalent to
CURLE_FTP_ACCEPT_FAILED in cURL error codes terms.
Instead, return CURLE_FAILED_INIT and output the failure reason as per
the other usage of getparameter().
To avoid the regression when users pass in passwords containing semi-
colons, we now drop the ability to set the login options with the same
options. Support for login options in CURLOPT_USERPWD was added in
7.31.0.
Test case 83 was modified to verify that colons and semi-colons can be
used as part of the password when using -u (CURLOPT_USERPWD).
Bug: http://curl.haxx.se/bug/view.cgi?id=1311
Reported-by: Petr Bahula
Assisted-by: Steve Holme
Signed-off-by: Daniel Stenberg <daniel@haxx.se>
Commit 32352ed6ad introduced various DNS options, however, these
would cause curl to exit with CURLE_NOT_BUILT_IN when c-ares wasn't
being used as the backend resolver even if the options weren't set
by the user.
Additionally corrected some minor coding style errors from the same
commit.
Moved the calls to checkpasswd() out of the getparameter() function
which allows for any related arguments to be specified on the command
line before or after --user (and --proxy-user).
For example: --bearer doesn't need to be specified before --user to
prevent curl from asking for an unnecessary password as is the case
with commit e7dcc454c6.
Added the ability to specify an XOAUTH2 bearer token [RFC6750] via the
--bearer option.
Example usage:
curl --url "imaps://imap.gmail.com:993/INBOX/;UID=1" --ssl-reqd
--bearer ya29.AHES6Z...OMfsHYI --user username@example.com
This function is meant to work *exactly* as curl_easy_perform() but will
use the event-based libcurl API internally instead of
curl_multi_perform(). To avoid relying on an actual event-based library
and to not use non-portable functions (like epoll or similar), there's a
rather inefficient emulation layer implemented on top of Curl_poll()
instead.
There's currently some convenience logging done in curl_easy_perform_ev
which helps when tracking down problems. They may be suitable to remove
or change once things seem to be fine enough.
curl has a new --test-event option when built with debug enabled that
then uses curl_easy_perform_ev() instead of curl_easy_perform(). If
built without debug, using --test-event will only output a warning
message.
NOTE: curl_easy_perform_ev() is not part if the public API on purpose.
It is only present in debug builds of libcurl and MUST NOT be considered
stable even then. Use it for libcurl-testing purposes only.
runtests.pl now features an -e command line option that makes it use
--test-event for all curl command line tests. The man page is updated.
A rather big overhaul and cleanup.
1 - curl wouldn't properly detect and reject globbing that ended with an
open brace if there were brackets or braces before it. Like "{}{" or
"[0-1]{"
2 - curl wouldn't properly reject empty lists so that "{}{}" would
result in curl getting (nil) strings in the output.
3 - By using strtoul() instead of sscanf() the code will now detected
over and underflows. It now also better parses the step argument to only
accept positive numbers and only step counters that is smaller than the
delta between the maximum and minimum numbers.
4 - By switching to unsigned longs instead of signed ints for the
counters, the max values for []-ranges are now very large (on 64bit
machines).
5 - Bumped the maximum number of globs in a single URL to 100 (from 10)
6 - Simplified the code somewhat and now it stores fixed strings as
single- entry lists. That's also one of the reasons why I did (5) as now
all strings between "globs" will take a slot in the array.
Added test 1234 and 1235 to verify. Updated test 87.
This commit fixes three separate bug reports.
Bug: http://curl.haxx.se/bug/view.cgi?id=1264
Bug: http://curl.haxx.se/bug/view.cgi?id=1265
Bug: http://curl.haxx.se/bug/view.cgi?id=1266
Reported-by: Will Dietz
Two fixes:
1. Force output file format to be stream-lf so that partial downloads
can be continued.
This should have minor impact as if the file does not exist, it was
created with stream-lf format. The only time this was an issue is if
there was already an existing file with a different format.
2. Fix file uploads are now fixed.
a. VMS binary files such as ZIP archives are now uploaded
correctly.
b. VMS text files are read once to get the correct size
and then converted to line-feed terminated records as
they are read into curl.
The default VMS text formats do not contain either line-feed or
carriage-return terminated records. Those delimiters are added by the
operating system file read calls if the application requests them.
Bug: http://curl.haxx.se/bug/view.cgi?id=496
We no longer pass our 'bool' data type variables nor constants as
an argument to my_setopt(), instead we use proper 1L or 0L values.
This also fixes macro used to pass string argument for CURLOPT_SSLCERT,
CURLOPT_SSLKEY and CURLOPT_EGDSOCKET using my_setopt_str() instead of
my_setopt().
This also casts enum or int argument data types to long when passed to
my_setopt_enum().
Implement wrappers around strtod to convert the user argument to a
double with sane error checking. Use this to allow --max-time and
--connect-timeout to accept decimal values instead of strictly integers.
The manpage is updated to make mention of this feature and,
additionally, forewarn that the actual timeout of the operation can
vary in its precision (particularly as the value increases in its
decimal precision).
The list of unsafe functions currently consists of sprintf, vsprintf,
strcat, strncat and gets.
Subsequently, some existing code needed updating to avoid warnings on
this.
The this_url pointer wasn't being initialized, so if strdup() would return
null when copying the filename in a metalink file, then hilarity would
ensue during the cleanup phase. This change was brought to you by clang,
which noticed this and raised a warning.
config_h.com is a new file that generates a config.h file based on the
curl_config.h.in file and a quick scan of the configure script. This is
actually a generic procedure that is shared with other VMS packages.
The existing pre-built config-vms.h had over 100 entries that were not
correct and in some cases conflicted with the build options available in
the build_vms.com.
generate_config_vms_h_curl.com is a helper procedure to the
config_h.com. It covers the cases that the generic config_h.com is not
able to figure out, and accepts input from the build_vms.com procedure.
build_curlbuild_h.com is a new file to generate the curlbuild.h file
that Curl is now using when it is using a curl_config.h file.
post-config-vms.h is a new file that is needed to provide VMS specific
definitions, and most of them need to be set before the system header
files are included.
The VMS build procedure is fixed:
1. Fixed to link in the correct HP ssl library.
2. Fixed to detect if HP Kerberos is installed.
3. Fixed to detect if HP LDAP is installed.
4. Fixed to detect if gnv$libzshr is installed.
5. Simplified the input parameter parsing to not use a loop.
6. Warn that 64 bit pointer option support is not complete
in comments.
7. Default to IEEE floating if platform supports it so
resulting libcurl will be compatible with other
open source projects on VMS.
8. Default to LARGEFILE if platform supports it.
9. Default to enable SSL, LDAP, Kerberos, libz
if the libraries are present.
10. Build with exact case global symbols for libcurl.
11. Generate linker option file needed.
12. Compiler list option only commonly needed items.
13. fulllist option for those who really want it.
14. Create debug symbol file on Alpha, IA64.
This reverts renaming and usage of lib/*.h header files done
28-12-2012, reverting 2 commits:
f871de0... build: make use of 76 lib/*.h renamed files
ffd8e12... build: rename 76 lib/*.h files
This also reverts removal of redundant include guard (redundant thanks
to changes in above commits) done 2-12-2013, reverting 1 commit:
c087374... curl_setup.h: remove redundant include guard
This also reverts renaming and usage of lib/*.c source files done
3-12-2013, reverting 3 commits:
13606bb... build: make use of 93 lib/*.c renamed files
5b6e792... build: rename 93 lib/*.c files
7d83dff... build: commit 13606bbfde follow-up 1
Start of related discussion thread:
http://curl.haxx.se/mail/lib-2013-01/0012.html
Asking for confirmation on pushing this revertion commit:
http://curl.haxx.se/mail/lib-2013-01/0048.html
Confirmation summary:
http://curl.haxx.se/mail/lib-2013-01/0079.html
NOTICE: The list of 2 files that have been modified by other
intermixed commits, while renamed, and also by at least one
of the 6 commits this one reverts follows below. These 2 files
will exhibit a hole in history unless git's '--follow' option
is used when viewing logs.
lib/curl_imap.h
lib/curl_smtp.h
Print "parsing (...) OK" only when no warnings are generated. If
no file is found in Metalink, treat it FAILED.
If no digest is provided, print WARNING in parse_metalink().
Also print validating FAILED after download.
These changes make tests 2012 to 2016 pass.
Including headers in response body will break Metalink XML parser.
If it is included in the file described in Metalink XML, hash check
will fail. Therefore, --include should be ignored if --metalink is
used.
The noprogress and isatty in Configurable are global, in a sense
that they persist in one curl invocation. Currently once one
download writes its response data to tty, they are set to FALSE
and they are not restored on successive downloads. This change
first backups the current noprogress and isatty, and restores
them when download does not write its data to tty.
In this change, --metalink option no longer takes argument. If
it is specified, given URIs are processed as Metalink XML file.
If given URIs are remote (e.g., http URI), curl downloads it
first. Regardless URI is local file (e.g., file URI scheme) or
remote, Metalink XML file is not written to local file system and
the received data is fed into Metalink XML parser directly. This
means with --metalink option, filename related options like -O
and -o are ignored.
Usage examples:
$ curl --metalink http://example.org/foo.metalink
This will download foo.metalink and parse it and then download
the URI described there.
$ curl --metalink file://foo.metalink
This will parse local file foo.metalink and then download the URI
described there.
Additionally, make hash checking ability mandatory in order to allow metalink
support in curl.
A command line option could be introduced to skip hash checking at runtime,
but the ability to check hashes should always be built-in when providing
metalink support.
Metalink file contains several hash types of checksums, such as
md5, sha-1, sha-256, etc. To deal with these checksums, I created
abstraction layer based on lib/curl_md5.h and
lib/md5.c. Basically, they are almost the same but I changed the
code so that it is not hash type dependent. Currently,
GNUTLS(nettle or gcrypt) and OpenSSL functions are supported.
Checksum checking is done by reopening download file. If there
is an I/O error, the current implementation just prints error
message and does not try next resource.
In this patch, the supported hash types are: md5, sha-1 and sha-256.
Filenames contained in Metalink file can include directory information.
Filenames are unique in Metalink file, taking into account the directory
information. So we need to create the directory hierarchy.
Curl has --create-dirs option, but we create directory hierarchy for
Metalink downloads regardless of the option value.
This patch also put metalink int variable outside of HAVE_LIBMETALINK
guard. This reduces the number of #ifdefs.
This change adds experimental Metalink support to curl.
To enable Metalink support, run configure with --with-libmetalink.
To feed Metalink file to curl, use --metalink option like this:
$ curl -O --metalink foo.metalink
We use libmetalink to parse Metalink files.
To achieve this, first new structure HeaderData is defined to hold
necessary data to perform header-related work. Then tool_header_cb now
receives HeaderData pointer as userdata. All header-related work
(currently, dumping header and Content-Disposition inspection) are done
in this callback function. HeaderData.outs->config is used to determine
whether each work is done.
Unit tests were also updated because after this change, curl code always
sets CURLOPT_HEADERFUNCTION and CURLOPT_HEADERDATA.
Tested with -O -J -D, -O -J -i and -O -J -D -i and all worked fine.
Explicit conversion to 'long' of curl_easy_setopt() third argument for options
CURLOPT_HTTPAUTH and CURLOPT_PROXYAUTH given that this is how its bitmasks are
docummented to be used.
Configuration files such as curl_config.h and all config-*.h no longer exist
nor are generated/copied into 'src' directory, now these only exist in 'lib'
directory from where curl tool sources uses them.
Additionally old src/setup.h has been refactored into src/tool_setup.h which
now pulls lib/setup.h
The possibility of a makefile needing an include path adjustment exists.
By modifying the parameter list for ourWriteOut() and passing the
OutStruct that collects data in tool_operate, we get access to the
remote name that we're writing to. Shell scripters should find this
useful when used in conjuntion with the --remote-header-name option.
This patch improves the output of curl's --libcurl option by
generating code which builds curl_httppost and curl_slist lists, and
uses symbolic names for enum and flag values. Variants of the
my_setopt macro in tool_setopt.h are added in order to pass extra type
information to the code-generation step in tool_setopt.c.
If curl is configured with --disable-libcurl-option then the macros
call curl_easy_setopt directly.
This new option tells curl to not work around a security flaw in the
SSL3 and TLS1.0 protocols. It uses the new libcurl option
CURLOPT_SSL_OPTIONS with the CURLSSLOPT_ALLOW_BEAST bit set.
Use the new library CURLOPT_TCP_KEEPALIVE rather than disabling this via
the sockopt callback. If --keepalive-time is used, apply the value to
CURLOPT_TCP_KEEPIDLE and CURLOPT_TCP_KEEPINTVL.
We want to continue to the next URL to try even on failures returned
from libcurl. This makes -f with ranges still get subsequent URLs even
if occasional ones return error. This was a regression as it used to
work and broke in the 7.23.0 release.
Added test case 1328 to verify the fix.
Bug: http://curl.haxx.se/bug/view.cgi?id=3481223
Reported by: Juan Barreto
A regression between 7.22.0 and 7.23.0 -- downloading a file with the
flags -O and -J results in the content being written to stdout if and
only if there was no Content-Disposition header in the http response. If
there is a C-D header with a filename attribute, the output is correctly
written.
Reported by: Dave Reisner
Bug: http://curl.haxx.se/mail/archive-2011-11/0030.html
Previously we required that -S/--show-error was used _after_
-s/--silent. This was slightly confusing since we strive to make
arguments as position independent as possible.
Now, you can use them in any order and the result should still be the
same.
Bug: http://curl.haxx.se/bug/view.cgi?id=3424286
Reported by: Andreas Olsson
