- do not grow memory by doubling its size
- do not leak previously allocated memory if reallocation fails
- replace while-loop with a single check to make sure
that the requested amount of data fits into the buffer
Bug: http://curl.haxx.se/bug/view.cgi?id=1450
Reported-by: Warren Menzer
There is no need to set the 'state' and 'result' member variables to
SMB_REQUESTING (0) and CURLE_OK (0) after the allocation via calloc()
as calloc() initialises the contents to zero.
I don't think both of my fix ups from yesterday were needed to fix the
compilation warning, so remove the one that I think is unnecessary and
let the next Android autobuild prove/disprove it.
smtp.c:2357 warning: adding 'size_t' (aka 'unsigned long') to a string
does not append to the string
smtp.c:2375 warning: adding 'size_t' (aka 'unsigned long') to a string
does not append to the string
smtp.c:2386 warning: adding 'size_t' (aka 'unsigned long') to a string
does not append to the string
Used array index notation instead.
This fixes compilation issues with compilers that don't support 64-bit
integers through long long or __int64 which was introduced in commit
07b66cbfa4.
Previously USE_NTLM2SESSION would only be defined automatically when
USE_NTRESPONSES wasn't already defined. Separated the two definitions
so that the user can manually set USE_NTRESPONSES themselves but
USE_NTLM2SESSION is defined automatically if they don't define it.
As the OpenSSL and NSS Crypto engines are prefered by the core NTLM
routines, to the Windows Crypt API, don't define USE_WIN32_CRYPT
automatically when either OpenSSL or NSS are in use - doing so would
disable NTLM2Session responses in NTLM type-3 messages.
If the scratch buffer was allocated in a previous call to
Curl_smtp_escape_eob(), a new buffer not allocated in the subsequent
call and no action taken by that call, then an attempt would be made to
try and free the buffer which, by now, would be part of the data->state
structure.
This bug was introduced in commit 4bd860a001.
Fixed a problem with the CRLF. detection when multiple buffers were
used to upload an email to libcurl and the line ending character(s)
appeared at the end of each buffer. This meant any lines which started
with . would not be escaped into .. and could be interpreted as the end
of transmission string instead.
This only affected libcurl based applications that used a read function
and wasn't reproducible with the curl command-line tool.
Bug: http://curl.haxx.se/bug/view.cgi?id=1456
Assisted-by: Patrick Monnerat
parsedate.c:548: warning: 'parsed' may be used uninitialized in this
function
As curl_getdate() returns -1 when parsedate() fails we can initialise
parsed to -1.
This fixes the test 506 torture test. The internal cookie API really
ought to be improved to separate cookie parsing errors (which may be
ignored) with OOM errors (which should be fatal).
As Windows based autoconf builds don't yet define USE_WIN32_CRYPTO
either explicitly through --enable-win32-cypto or automatically on
_WIN32 based platforms, subsequent builds broke with the following
error message:
"Can't compile NTLM support without a crypto library."
Fixed an issue with the message size calculation where the raw bytes
from the buffer were interpreted as signed values rather than unsigned
values.
Reported-by: Gisle Vanem
Assisted-by: Bill Nagel
Don't use a hard coded size of 4 for the security layer and buffer size
in Curl_sasl_create_gssapi_security_message(), instead, use sizeof() as
we have done in the sasl_gssapi module.
Reduced the amount of free's required for the decoded challenge message
in Curl_sasl_create_gssapi_security_message() as a result of coding it
differently in the sasl_gssapi module.
Sending NTLM/Negotiate header again after successful authentication
breaks the connection with certain Proxies and request types (POST to MS
Forefront).
The ability to do HTTP requests over a UNIX domain socket has been
requested before, in Apr 2008 [0][1] and Sep 2010 [2]. While a
discussion happened, no patch seems to get through. I decided to give it
a go since I need to test a nginx HTTP server which listens on a UNIX
domain socket.
One patch [3] seems to make it possible to use the
CURLOPT_OPENSOCKETFUNCTION function to gain a UNIX domain socket.
Another person wrote a Go program which can do HTTP over a UNIX socket
for Docker[4] which uses a special URL scheme (though the name contains
cURL, it has no relation to the cURL library).
This patch considers support for UNIX domain sockets at the same level
as HTTP proxies / IPv6, it acts as an intermediate socket provider and
not as a separate protocol. Since this feature affects network
operations, a new feature flag was added ("unix-sockets") with a
corresponding CURL_VERSION_UNIX_SOCKETS macro.
A new CURLOPT_UNIX_SOCKET_PATH option is added and documented. This
option enables UNIX domain sockets support for all requests on the
handle (replacing IP sockets and skipping proxies).
A new configure option (--enable-unix-sockets) and CMake option
(ENABLE_UNIX_SOCKETS) can disable this optional feature. Note that I
deliberately did not mark this feature as advanced, this is a
feature/component that should easily be available.
[0]: http://curl.haxx.se/mail/lib-2008-04/0279.html
[1]: http://daniel.haxx.se/blog/2008/04/14/http-over-unix-domain-sockets/
[2]: http://sourceforge.net/p/curl/feature-requests/53/
[3]: http://curl.haxx.se/mail/lib-2008-04/0361.html
[4]: https://github.com/Soulou/curl-unix-socket
Signed-off-by: Peter Wu <peter@lekensteyn.nl>
On some platforms curl would crash if no credentials were used. As such
added detection of such a use case to prevent this from happening.
Reported-by: Gisle Vanem
This patch prepares for adding UNIX domain sockets support.
TCP_NODELAY and TCP_KEEPALIVE are specific to TCP/IP sockets, so do not
apply these to other socket types. bindlocal only works for IP sockets
(independent of TCP/UDP), so filter that out too for other types.
Signed-off-by: Peter Wu <peter@lekensteyn.nl>
smb.c:398: warning: comparison of integers of different signs:
'ssize_t' (aka 'long') and 'unsigned long'
smb.c:443: warning: comparison of integers of different signs:
'ssize_t' (aka 'long') and 'unsigned long'
smb.c:322: warning: conversion to 'short unsigned int' from 'unsigned
int' may alter its value
smb.c:323: warning: conversion to 'short unsigned int' from 'unsigned
int' may alter its value
smb.c:482: warning: conversion to 'short unsigned int' from 'int' may
alter its value
smb.c:521: warning: conversion to 'unsigned int' from 'curl_off_t' may
alter its value
smb.c:549: warning: conversion to 'unsigned int' from 'curl_off_t' may
alter its value
smb.c:550: warning: conversion to 'short unsigned int' from 'int' may
alter its value
smb.c:489: warning: declaration of 'close' shadows a global declaration
smb.c:511: warning: declaration of 'read' shadows a global declaration
smb.c:528: warning: declaration of 'write' shadows a global declaration
smb.c:212: warning: unused parameter 'done'
smb.c:380: warning: ISO C does not allow extra ';' outside of a function
smb.c:812: warning: unused parameter 'premature'
smb.c:822: warning: unused parameter 'dead'
smb.c:311: warning: conversion from 'unsigned __int64' to 'u_short',
possible loss of data
smb.c:425: warning: conversion from '__int64' to 'unsigned short',
possible loss of data
smb.c:452: warning: conversion from '__int64' to 'unsigned short',
possible loss of data
smb.c:162: error: comma at end of enumerator list
smb.c:469: warning: conversion from 'size_t' to 'unsigned short',
possible loss of data
smb.c:517: warning: conversion from 'curl_off_t' to 'unsigned int',
possible loss of data
smb.c:545: warning: conversion from 'curl_off_t' to 'unsigned int',
possible loss of data
If the scratch buffer already existed when the CRLF conversion was
performed then the buffer pointer would be checked twice for NULL. This
second check is only necessary if the call to malloc() was performed by
the first check.
Whilst I had moved the dot stuffing code from being performed before
CRLF conversion takes place to after it, in commit 4bd860a001, I had
moved it outside the 'when something read' block of code when meant
it could perform the dot stuffing twice on partial send if nread
happened to contain the right values. It also meant the function could
potentially read past the end of buffer. This was highlighted by the
following warning:
warning: `nread' might be used uninitialized in this function
