Curl_getconnectinfo() is changed to return a proper curl_socket_t for
the last socket so that it'll work more portably (and cause less
compiler warnings).
Add a timeout check for handles in the state machine so that they will
timeout in all states disregarding what actions that may or may not
happen.
Fixed a bug in socket_action introduced recently when looping over timed
out handles: it wouldn't assign the 'data' variable and thus it wouldn't
properly take care of handles.
In the update_timer function, the code now checks if the timeout has
been removed and then it tells the application. Previously it would
always let the remaining timeout(s) just linger to expire later on.
Each easy handle has a list of timeouts, so as soon as the main timeout
for a handle expires, we must make sure to get the next entry from the
list and re-add the handle to the splay tree.
This was attempted previously but was done poorly in my commit
232ad6549a.
When a new transfer is about to start we now set the proper timeouts to
expire for the multi interface if they are set for the handle. This is a
follow-up bugfix to make sure that easy handles timeout properly when
the times expire and the multi interface is used. This also improves
curl_multi_timeout().
The fix for the busyloop really only is a temporary work-around. It
causes a BLOCKING behavior which is a NO-NO. This function should rather
be split up in a do and a doing piece where the pieces that aren't
possible to send now will be sent in the doing function repeatedly until
the entire request is sent.
HTTP allows that a server sends trailing headers after all the chunks
have been sent WITHOUT signalling their presence in the first response
headers. The "Trailer:" header is only a SHOULD there and as we need to
handle the situation even without that header I made libcurl ignore
Trailer: completely.
Test case 1116 was added to verify this and to make sure we handle more
than one trailer header properly.
Reported by: Patrick McManus
Bug: http://curl.haxx.se/bug/view.cgi?id=3052450
The script works exactly same as the Perl one except for one thing:
when the text descriptions generated with openssl are included then
the md5 fingerprints are missing; seems openssl has either a bug or
a feature which prints the md5 fingerprint output to stdout instead
of writing them to specified file; this script could here do the same
as what the Perl scripr does (redirect stdout into file) but this
makes the script take up double the time because it needs to launch
cmd.exe 140 times (fo each openssl call). So I think for now we just
ommit the md5 fingerprints, and see if openssl will be fixed.
I fell over this bug report that mentioned that libcurl could wrongly
send more than one complete messages at the end of a transfer. Reading
the code confirmed this, so I've added a new multi state to make it not
happen. The mentioned bug report was made by Brad Jorsch but is (oddly
enough) filed in Debian's bug tracker for the "wmweather+" tool.
Bug: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=593390
There's an error in http_negotiation.c where a mistake is using only
userpwd even for proxy requests. Ludek provided a patch, but I decided
to write the fix slightly different using his patch as inspiration.
Reported by: Ludek Finstrle
Bug: http://curl.haxx.se/bug/view.cgi?id=3046066
When detecting that the send or recv speed, the multi interface changes
state to TOOFAST and previously there was no timeout set that would
force a recheck but it would rely on the application to somehow call
libcurl anyway. This now sets a timeout for a suitable future time to
check again if the average transfer speed is then below the threshold
again.
Curl_expire() is now expanded to hold a list of timeouts for each easy
handle. Only the closest in time will be the one used as the primary
timeout for the handle and will be used for the splay tree (which sorts
and lists all handles within the multi handle).
When the main timeout has triggered/expired, the next timeout in time
that is kept in the list will be moved to the main timeout position and
used as the key to splay with. This way, all timeouts that are set with
Curl_expire() internally will end up as a proper timeout. Previously any
Curl_expire() that set a _later_ timeout than what was already set was
just silently ignored and thus missed.
Setting Curl_expire() with timeout 0 (zero) will cancel all previously
added timeouts.
Corrects known bug #62.
Instead of looping over all attached easy handles, this now keeps a list
of messages in the multi handle. It allows curl_multi_info_read() to
perform O(1) no matter how many easy handles that are handled. This is
of importance since this function may be polled very frequently by apps
using the multi interface.
When the progress callback is called during the TCP connection, an error
return would accidentally not abort the operation as intended but would
instead be counted as a failure to connect to that particular IP and
libcurl would just continue to try the next. I made singleipconnect()
and trynextip() return CURLcode properly.
Added bonus: it corrected the error code for bad --interface usages,
like tested in test 1084 and test 1085.
Reported by: Adam Light
Bug: http://curl.haxx.se/mail/lib-2010-08/0105.html
Added the -br switch to dynamic builds which fixes the issue I saw
with curl's --version output. Added debug info and symfile for debug
builds to linker opts. Added DLL loader for wlink back, but this time
dependend on wlink version.
Patch posted to the list by malak.jiri AT gmail.com.
The var %MAKEFLAGS is only set in 3 cases: if set as environment
var or as macro definition from commandline, and either with the
-u or -ms switch. Since all these cases are unlikely for the average
user it should be safe to only test if %MAKEFLAGS is defined; this
has the benefit that now all other switches can be used again in
addition to the -u which was formerly not possible.
Curl_llist_init is never used outside of llist.c and thus it should be
static. I also removed the protos for Curl_llist_insert_prev and
Curl_llist_remove_next which are functions we removed from llist.c ages
ago.
Test 563 is enabled now and verifies that the combo FTP type=A URL,
CURLOPT_PORT set and proxy work fine. As a bonus I managed to remove the
somewhat odd FTP check in parse_remote_port() and instead converted it
to a better and more generic 'slash_removed' struct field. Checking the
->protocol field isn't right since when an FTP:// URL is sent over a
HTTP proxy, the protocol is HTTP but the URL was handled by the FTP code
and thus slash_removed is set TRUE for this case.
The struct used for storing the message for a completed transfer is now
no longer allocated separatly but is kept within the main struct kept
for each easy handle so that we avoid one malloc (and the subsequent
free).
When libcurl internally decided to wait for a 100-continue header, there
was no call to the timeout function so there was no timeout callback
called when the multi_socket API was used and thus applications became
either completely wrong or at least ineffecient depending on how they
handled the situation. We now set a timeout to get triggered.
Reported by: Ben Darnell
Bug: http://curl.haxx.se/bug/view.cgi?id=3039744
libssh2 1.2.6 and later handle >32bit file sizes properly even on 32bit
architectures and we make sure to use that ability.
Reported by: Mikael Johansson
Bug: http://curl.haxx.se/mail/lib-2010-08/0052.html
Simply because the TCP might be connected already we cannot skip the
proxy connect procedure. We need to be careful to not overload more
meaning to the bits.tcpconnect field like this.
With this fix, SOCKS proxies work again when the multi interface is
used. I believe this regression was added with commit 4b351d018e,
released as 7.20.1.
Left todo: add a test case that verifies this functionality that
prevents us from breaking it again in the future!
Reported by: Robin Cornelius
Bug: http://curl.haxx.se/bug/view.cgi?id=3033966
Commit 496002ea1c (released in 7.20.1) broke FTPS when using the
multi interface and OpenSSL was used. The condition for the non-blocking
connect was incorrect.
Reported by: Georg Lippitsch
Bug: http://curl.haxx.se/mail/lib-2010-07/0270.html
Previously the host name buffer was only used if gethostname() exists,
but since we converted that into a curl private function that function
always exists and will be used so the buffer needs to exist for all
cases/systems.
A shared library tests/libtest/.libs/lihostname.so is preloaded in NTLM
test-cases to override the system implementation of gethostname(). It
makes it possible to test the NTLM authentication for exact match, and
this way test the implementation of MD4 and DES.
If LD_PRELOAD doesn't work, a debug build willl also workk as debug
builds are now made to prefer a specific environment variable and will
then return that content as host name instead of the actual one.
Kamil wrote the bulk of this, Daniel Stenberg polished it.
lib/Makefile.Watcom works fine already, for src/Makefile.Watcom we
need first to tweak src/Makefile.inc a bit - therefore the handtweaked
list still exists for now.
- make both libcurl and curl makefiles use register calling convention
(previously libcurl had stack calling convention).
- added include paths to the Watcom headers so its no longer required
to set the environment vars for this.
- added -wcd=201 to supress compiler warning about unreachable code.
- use macros for all tools, and removed dependency on GNU tools like rm.
- make ipv6 and debug builds controlable via env vars and so make them
optional instead of default.
- commented WINLDAPAPI and WINBERAPI since they broke with OW 1.8, and
it seems they're not needed (anymore?).
- added rule for hugehelp.c.cvs so that it will be created when not
already exist - this is required for building from a release tarball
since there we have no hugehelp.c.cvs, thus compilation broke.
- removed C_ARG creation from lib/Makefile.Watcom and use CFLAGS
directly as done too in src/Makefile.Watcom - this has the benefit
that we will see all active cflags and defines during compile.
- added LINK-ARG to src/Makefile.Watcom in order to better control
linker input.
- a couple of other minor makefile tweaks here and there ...
- added largefile support for Watcom builds to config-win32.h. Not yet
tested if it really works, but should since Win32 supports it.
- added loaddll stuff to speed up builds if supported.
Win64's 32 bit long but 64 bit size_t caused a warning that we avoid
with a typecast. A small whitespace indent fix was also applied.
Reported by: Adam Light
This passes -Werror to gcc when building curl and libcurl,
allowing easy dection of compile warnings.
Signed-off-by: Ben Greear <greearb@candelatech.com>
... since FTP is using it as well, and potentially other protocols!
Also, an #endif CURL_DISABLE_HTTP was incorrectly marked, as it seems to
end the proxy block instead.
The FTP implementation was missing a timestamp reset point, making the
waiting for responses after sending a post-transfer "QUOTE" command not
working as supposedly. This bug was introduced in 7.20.0
curl_multi perform has two phases: run through every easy handle calling
multi_runsingle and remove expired timers (timer removal).
If a small timer (e.g. 1-10ms) is set during multi_runsingle, then it's
possible that the timer has passed by when the timer removal runs. The
timer which was just added is then removed. This will potentially cause
the timer list to be empty and cause the next call to curl_multi_timeout
to return -1. Ideally, curl_multi_timeout should return 0 in this case.
One way to fix this is to move the struct timeval now = Curl_tvnow(); to
the top of curl_multi_perform. The change does that.
As mentioned in bug report #2956968, the HTTP code wouldn't send the
first empty chunk during the auth negotiation phase of the HTTP request
sending, so the server would wait for data to come and libcurl would
wait for data to arrive... I've made the code not enable chunked
encoding until the auth negotiation is done and thus this scenario
doesn't occur anymore.
Reported by: Sidney San Martn
Bug: http://curl.haxx.se/bug/view.cgi?id=2956968
When curl_multi_remove_handle() is called and an easy handle is returned
to the connection cache held in the multi handle, then we cannot allow
CURLINFO_LASTSOCKET to extract it since that will more or less encourage
that the user uses the socket while it can get used by libcurl again.
Without this fix, we'd get a segfault in Curl_getconnectinfo() trying to
dereference the NULL pointer in 'data->state.connc'.
Bug: http://curl.haxx.se/bug/view.cgi?id=3023840
When configured with '--without-ssl --with-nss', NTLM authentication
now uses NSS crypto library for MD5 and DES. For MD4 we have a local
implementation in that case. More details are available at
https://bugzilla.redhat.com/603783
In order to get it working, curl_global_init() must be called with
CURL_GLOBAL_SSL or CURL_GLOBAL_ALL. That's necessary because NSS needs
to be initialized globally and we do so only when the NSS library is
actually required by protocol. The mentioned call of curl_global_init()
is responsible for creating of the initialization mutex.
There was also slightly changed the NSS initialization scenario, in
particular, loading of the NSS PEM module. It used to be loaded always
right after the NSS library was initialized. Now the library is
initialized as soon as any SSL or NTLM is required, while the PEM module
is prevented from being loaded until the SSL is actually required.
When a hostname resolves to multiple IP addresses and the first one
tried doesn't work, the socket for the second attempt may get dropped on
the floor, causing the request to eventually time out. The issue is that
when using kqueue (as on mac and bsd platforms) instead of select, the
kernel removes the first fd from kqueue when it is closed (in trynextip,
connect.c:503). Trynextip() then goes on to open a new socket, which
gets assigned the same number as the one it just closed. Later in
multi.c, socket_cb is not called because the fd is already in
multi->sockhash, so the new socket is never added to kqueue.
The correct fix is to ensure that socket_cb is called to remove the fd
when trynextip() closes the socket, and again to re-add it after
singleipsocket(). I'm not sure how to cleanly do that, but the attached
patch works around the problem in an admittedly kludgy way by delaying
the close to ensure that the newly-opened socket gets a different fd.
Daniel's added comment: I didn't spot a way to easily do a nicer fix so
I've proceeded with Ben's patch.
Bug: http://curl.haxx.se/bug/view.cgi?id=3017819
Patch by: Ben Darnell
For example the libssh2 based functions return other negative
values than -1 to signal errors and it is important that we catch
them properly. Right before this, various failures from libssh2
were treated as negative download amounts which caused havoc.
My additional call to Curl_pgrsUpdate() would sometimes get
called even though there's no connection (left) so a NULL pointer
would get passed, causing a segfault.
1) no need to call the progress function twice when in the
CURLM_STATE_TOOFAST state.
2) Make sure that the progress callback's return code is
acknowledged when used
As long as no error is reported, the progress function can get
called. This may be a little TOO often so we should keep an eye
on this and possibly make this conditional somehow.
Older unixes want an 'int' instead of 'size_t' as the 3rd
argumment so before this change it would cause warnings such as:
There is an implicit conversion from "unsigned long" to "int";
rounding, sign extension, or loss of accuracy may result.
Was seeing spurious SSL connection aborts using libcurl and
OpenSSL. I tracked it down to uncleared error state on the
OpenSSL error stack - patch attached deals with that.
Rough idea of problem:
Code that uses libcurl calls some library that uses OpenSSL but
don't clear the OpenSSL error stack after an error.
ssluse.c calls SSL_read which eventually gets an EWOULDBLOCK from
the OS. Returns -1 to indicate an error
ssluse.c calls SSL_get_error. First thing, SSL_get_error calls
ERR_get_error to check the OpenSSL error stack, finds an old
error and returns SSL_ERROR_SSL instead of SSL_ERROR_WANT_READ or
SSL_ERROR_WANT_WRITE.
ssluse.c returns an error and aborts the connection
Solution:
Clear the openssl error stack before calling SSL_* operation if
we're going to call SSL_get_error afterwards.
Notes:
This is much more likely to happen with multi because it's easier
to intersperse other calls to the OpenSSL library in the same
thread.
Enable OpenLDAP support for cygwin builds. This support was disabled back
in 2008 due to incompatibilities between OpenSSL and OpenLDAP headers.
cygwin's OpenSSL 0.9.8l and OpenLDAP 2.3.43 versions on cygwin 1.5.25
allow building an OpenLDAP enabled libcurl supporting back to Windows 95.
Remove non-functional CURL_LDAP_HYBRID code and references.
Jason McDonald posted bug report #3006786 when he found that the
SFTP code didn't timeout properly in several places in the code
even if a timeout was set properly.
Based on his suggested patch, I wrote a different implementation
that I think addressed the issue better and also uses the connect
timeout for the initial part of the SSH/SFTP done during the
"protocol connect" phase.
(http://curl.haxx.se/bug/view.cgi?id=3006786)
Igor Novoseltsev reported a problem with the multi socket API and
using timeouts and timers. It boiled down to a problem with
libcurl's use of GetTickCount() interally to figure out the
current time, while Igor's own application code used another
function call.
It made his app call the socket API timeout function a bit
_before_ libcurl would consider the timeout to trigger, and that
could easily lead to timeouts or stalls in the app. It seems
GetTickCount() in general often has no better resolution than
16ms and switching to the alternative function
QueryPerformanceCounter has its share of problems:
http://www.virtualdub.org/blog/pivot/entry.php?id=106
We address this problem by simply having libcurl treat timers
that already has occured or will occur within 40ms subject for
treatment. I'm confident that there are other implementations and
operating systems with similarly in accurate timer functions so
it makes sense to have applied generically and I don't believe we
sacrifice much by adding a 40ms inaccuracy on these timeouts.
