]> asedeno.scripts.mit.edu Git - PuTTY.git/log
PuTTY.git
11 years agoGive a better error message if a PuTTY private key file has a version
Simon Tatham [Fri, 22 Feb 2013 21:39:02 +0000 (21:39 +0000)]
Give a better error message if a PuTTY private key file has a version
number we don't understand. It's nicer to report 'format too new' than
'not a PuTTY key file'.

[originally from svn r9763]

11 years agoAdd support for HMAC-SHA-256 as an SSH-2 MAC algorithm ("hmac-sha2-256")
Ben Harris [Wed, 20 Feb 2013 23:30:55 +0000 (23:30 +0000)]
Add support for HMAC-SHA-256 as an SSH-2 MAC algorithm ("hmac-sha2-256")
as specified in RFC 6668.  This is not so much because I think it's
necessary, but because scrypt uses HMAC-SHA-256 and once we've got it we
may as well use it.

Code very closely derived from the HMAC-SHA-1 code.

Tested against OpenSSH 5.9p1 Debian-5ubuntu1.

[originally from svn r9759]

11 years agoTake advantage of PUT_32BIT_MSB_FIRST when constructing sequence numbers
Ben Harris [Wed, 20 Feb 2013 22:37:34 +0000 (22:37 +0000)]
Take advantage of PUT_32BIT_MSB_FIRST when constructing sequence numbers
to MAC.

[originally from svn r9758]

11 years agoPatch from Egmont Koblinger to implement two extended variants of
Simon Tatham [Wed, 23 Jan 2013 22:59:17 +0000 (22:59 +0000)]
Patch from Egmont Koblinger to implement two extended variants of
xterm mouse tracking, both supported by the current up-to-date xterm
(288). They take the form of two new DEC terminal modes, 1006 and
1015, which do not in themselves _enable_ mouse tracking but they
modify the escape sequences sent if mouse tracking is enabled in the
usual way.

[originally from svn r9752]

11 years agoAdjust comments around split_into_argv() to clarify that it's not
Simon Tatham [Sat, 19 Jan 2013 17:17:44 +0000 (17:17 +0000)]
Adjust comments around split_into_argv() to clarify that it's not
*Windows's* command-line splitting rules we're mimicking here; it's
VC7's, and they're not the same as VC10's.

[originally from svn r9748]

11 years agoReorganise setup_fonts_ucs so that in case of error it does nothing
Simon Tatham [Sun, 13 Jan 2013 21:59:10 +0000 (21:59 +0000)]
Reorganise setup_fonts_ucs so that in case of error it does nothing
and returns its error message as a string, instead of actually
printing it on standard error and exiting. Now we can preserve the
previous error behaviour when we get a nonexistent font name at
startup time, but no longer rudely terminate in mid-session if the
user configures a bogus font name in Change Settings.

[originally from svn r9745]

11 years agoMove the -I options for all our source subdirectories out of AM_CFLAGS
Simon Tatham [Mon, 7 Jan 2013 19:08:33 +0000 (19:08 +0000)]
Move the -I options for all our source subdirectories out of AM_CFLAGS
and into AM_CPPFLAGS. This is more conceptually sensible according to
my reading of the automake manual, and also has the specific desirable
effect that they move to the front of the command line, ahead of any
'system' type -I options that autoconf might have felt a need for.

A user reported that autoconf had added -I/usr/local/include to their
command line for the sake of a required header file, but their
/usr/local/include also turned out to include a thing called 'proxy.h'
(from libproxy, nothing to do with us) which shadowed our own proxy.h
and caused a build failure. This should fix that.

[originally from svn r9736]

11 years agoUpdate the suggested compile command in sshbn.c's test rig.
Simon Tatham [Sat, 22 Dec 2012 18:10:08 +0000 (18:10 +0000)]
Update the suggested compile command in sshbn.c's test rig.

[originally from svn r9732]

11 years agoFix two gcc warnings about confused printf format strings in the
Simon Tatham [Sat, 22 Dec 2012 18:09:02 +0000 (18:09 +0000)]
Fix two gcc warnings about confused printf format strings in the
bignum code's test harness. Thanks to Sup Yut Sum for fixing this in
TortoisePlink and Sven Strickroth for bringing it to my attention.

[originally from svn r9731]

11 years agoRonald Landheer-Cieslak points out that the various back ends which
Simon Tatham [Sat, 22 Dec 2012 09:40:47 +0000 (09:40 +0000)]
Ronald Landheer-Cieslak points out that the various back ends which
treat all socket closures as clean exits (because the protocol doesn't
provide for transferring a process exit code) could usefully at least
treat _socket errors_ as unclean exits. Patch the Telnet, Rlogin and
Raw backends to retain that information and return INT_MAX to the
frontend.

I wasn't sure whether it was better to solve this by modifying each
affected frontend, or each affected backend. I chose the latter, but
neither is really ideal; this is the sort of thing that makes me wish
we had a piece of fixed middleware in between, independent of both
platform and protocol.

[originally from svn r9730]

11 years agoSwitch round a bogus if statement I've just noticed. Both the write to
Simon Tatham [Tue, 18 Dec 2012 09:19:04 +0000 (09:19 +0000)]
Switch round a bogus if statement I've just noticed. Both the write to
pty_utmp_helper_pipe _and_ the close of it if we're not going to write
should be conditionalised on the pipe existing, rather than just the
former!

[originally from svn r9729]

11 years agoPatch from Brad Smith to use posix_openpt() instead of
Simon Tatham [Tue, 18 Dec 2012 09:02:38 +0000 (09:02 +0000)]
Patch from Brad Smith to use posix_openpt() instead of
open("/dev/ptmx"), where the former is available. Improves
portability, since at least one OS (OpenBSD) supports the POSIX pty
functions but does it via an underlying mechanism which doesn't
involving having a /dev/ptmx.

[originally from svn r9728]

11 years agoUse O_NOCTTY (if available) when opening /dev/ptmx, just in case any
Simon Tatham [Tue, 18 Dec 2012 09:02:38 +0000 (09:02 +0000)]
Use O_NOCTTY (if available) when opening /dev/ptmx, just in case any
OS doesn't automatically assume it.

(It would seem faintly weird to me - surely opening the master end of
a given pty is a fairly good indication that you're _not_ a process
running inside it which wants to have it available as /dev/tty! But
you never know...)

[originally from svn r9727]

11 years agoAdd a bounds check in the word-by-word selection code to prevent
Simon Tatham [Mon, 17 Dec 2012 08:40:33 +0000 (08:40 +0000)]
Add a bounds check in the word-by-word selection code to prevent
attempting to call lineptr() with a y-coordinate off the bottom of the
screen and triggering the dreaded 'line==NULL' message box.

This crash can only occur if the bottommost line of the screen has the
LATTR_WRAPPED flag set, which as far as I can see you can only
contrive by constructing a LATTR_WRAPPED line further up the screen
and then moving it down using an insert-line escape sequence. That's
probably why this bug has been around forever without anyone coming
across it.

[originally from svn r9726]

11 years agoPatch from Hideki Eiraku to make PuTTY call GetScrollInfo, so it can
Simon Tatham [Tue, 4 Dec 2012 20:53:19 +0000 (20:53 +0000)]
Patch from Hideki Eiraku to make PuTTY call GetScrollInfo, so it can
use 32-bit scrollbar position data instead of being limited to the
16-bit version that comes in scrollbar messages' wParam.

[originally from svn r9720]

11 years agoFix another error-reporting bug, in which sk_newlistener would fail to
Simon Tatham [Wed, 14 Nov 2012 18:32:09 +0000 (18:32 +0000)]
Fix another error-reporting bug, in which sk_newlistener would fail to
capture the error code if listen() returned an error, and instead pass
0 (saved from the previous successful bind) to winsock_error_string.

[originally from svn r9708]

11 years agoWhen manually initialising a 'struct RSAKey' due to loading an SSH1
Simon Tatham [Tue, 13 Nov 2012 21:34:12 +0000 (21:34 +0000)]
When manually initialising a 'struct RSAKey' due to loading an SSH1
public key but not the private half, NULL out all the CRT-optimisation
fields as well as the private exponent pointer. Otherwise segfaults -
security-harmless, but annoying - can happen in freersakey() when we
notice they aren't null and try to free them.

[originally from svn r9705]

11 years agoAdd a fallback case to winsock_error_string() which makes it call
Simon Tatham [Tue, 13 Nov 2012 18:36:27 +0000 (18:36 +0000)]
Add a fallback case to winsock_error_string() which makes it call
FormatMessage to get the OS's text for any error not in our own
translation table. Should eliminate the frustrating 'unknown error'.

(I haven't chosen to use FormatMessage unconditionally, because it
comes out with enormous messages along the lines of "No connection
could be made because the target machine actively refused it" in place
of "Connection refused" and I'm Unixy enough to prefer the latter.
Also, on older Windowses, Winsock error codes are in a separate API
segment and don't work with FormatMessage anyway.)

[originally from svn r9704]

11 years agoWindows's sk_address_is_local() was returning the wrong answers for
Simon Tatham [Wed, 17 Oct 2012 20:48:07 +0000 (20:48 +0000)]
Windows's sk_address_is_local() was returning the wrong answers for
IPv6 addresses, because I'd mistakenly cast an ai_addr to the low-
level 'struct in6_addr' instead of the correct 'struct sockaddr_in6'.

[originally from svn r9690]

11 years agoWhen a proxy negotiation function is called with PROXY_CHANGE_NEW, it
Simon Tatham [Tue, 16 Oct 2012 20:15:52 +0000 (20:15 +0000)]
When a proxy negotiation function is called with PROXY_CHANGE_NEW, it
should not call plug functions, because it's being called from within
new_connection(), and the state on which the plug functions depend
will not have been set up until new_connection() returns success.
Instead, we set the error string in the Proxy_Socket, which will cause
the same error message to be returned as a failure of new_connection().

[originally from svn r9689]

11 years agoIf you configure Unix PuTTY to use a proxy, tell it to even proxy
Simon Tatham [Tue, 16 Oct 2012 20:15:51 +0000 (20:15 +0000)]
If you configure Unix PuTTY to use a proxy, tell it to even proxy
localhost connections, and also enable X forwarding in such a way that
it will attempt to connect to a Unix-domain X server socket, an
assertion will fail when proxy_for_destination() tries to call
sk_getaddr(). Fix by ensuring that Unix-domain sockets are _never_
proxied, since they fundamentally can't be.

[originally from svn r9688]

11 years agoAdd timestamps to the 'SSH raw data' logging mode.
Simon Tatham [Wed, 10 Oct 2012 18:32:23 +0000 (18:32 +0000)]
Add timestamps to the 'SSH raw data' logging mode.

[originally from svn r9687]

11 years agoA user points out that we should free the 'hProcess' and 'hThread'
Simon Tatham [Wed, 10 Oct 2012 18:29:16 +0000 (18:29 +0000)]
A user points out that we should free the 'hProcess' and 'hThread'
handles returned in the PROCESS_INFORMATION structure after we call
CreateProcess.

[originally from svn r9686]

11 years agoSumudu Fernando points out that in the big r9214 destabilisation I
Simon Tatham [Tue, 2 Oct 2012 19:31:33 +0000 (19:31 +0000)]
Sumudu Fernando points out that in the big r9214 destabilisation I
mistakenly rearranged the logic in an if statement in window.c, with
the effect that scroll-wheel events are no longer sent via xterm mouse
tracking. Put it back to the way it was.

[originally from svn r9679]
[r9214 == a1f3b7a358adaa7c2a98359cd0373aa823eeb14b]

11 years agoClip the 'lines' parameter to scroll() at the size of the scroll
Simon Tatham [Sun, 23 Sep 2012 15:36:54 +0000 (15:36 +0000)]
Clip the 'lines' parameter to scroll() at the size of the scroll
window. scroll() iterates that many times, so this prevents a tedious
wait if you give a very large parameter to ESC[L or ESC[M, for
example.

A side effect is that very large requests for upward scrolling in a
context that affects the scrollback will not actually wipe out the
whole scrollback: instead they push just the current lines of the
screen into the scrollback, and don't continue on to fill it up with
endless boring blank lines. I think this is likely to be more useful
in general, since it avoids wiping out lots of useful scrollback data
by mistake. I can imagine that people might have been using it
precisely _to_ wipe the scrollback in some situations, but if so then
they should use CSI 3 J instead.

[originally from svn r9677]

11 years agoWindows PSFTP has two places that call run_timers and I'd only updated
Ben Harris [Wed, 19 Sep 2012 22:17:10 +0000 (22:17 +0000)]
Windows PSFTP has two places that call run_timers and I'd only updated
one of them.  Correct that.

[originally from svn r9676]

11 years agoFix indentation mess in my timing overhaul.
Ben Harris [Wed, 19 Sep 2012 22:16:30 +0000 (22:16 +0000)]
Fix indentation mess in my timing overhaul.

[originally from svn r9675]

11 years agoFix a stupid sign bug in run_timers() that broke Windows Plink (and
Ben Harris [Wed, 19 Sep 2012 22:12:00 +0000 (22:12 +0000)]
Fix a stupid sign bug in run_timers() that broke Windows Plink (and
should really have broken everything else too).

[originally from svn r9674]

11 years agoMake --help and --version work consistently across all tools.
Simon Tatham [Wed, 19 Sep 2012 17:08:15 +0000 (17:08 +0000)]
Make --help and --version work consistently across all tools.

Well, at least across all command-line tools on both Windows and Unix,
and the GTK apps on Unix too. The Windows GUI apps fundamentally can't
write to standard output and it doesn't seem sensible to use message
boxes for these purposes :-)

[originally from svn r9673]

11 years agoBetter document the various environments with which Makefile.cyg works.
Ben Harris [Tue, 18 Sep 2012 23:07:42 +0000 (23:07 +0000)]
Better document the various environments with which Makefile.cyg works.

[originally from svn r9670]

11 years agoAdd a new COMPAT option for environments lacking SecureZeroMemory(),
Ben Harris [Tue, 18 Sep 2012 23:05:29 +0000 (23:05 +0000)]
Add a new COMPAT option for environments lacking SecureZeroMemory(),
rather than explicitly checking for Winelib.  It seems that w32api is
lacking it as well.

[originally from svn r9669]

11 years agoAllow remote-to-local forwardings to use IPv6.
Ben Harris [Tue, 18 Sep 2012 21:50:47 +0000 (21:50 +0000)]
Allow remote-to-local forwardings to use IPv6.

RFC 4245 section 7.1 specifies the meaning of the "address to bind"
parameter in a "tcpip-forward" request.  "0.0.0.0" and "127.0.0.1" are
specified to be all interfaces and the loopback interface respectively
in IPv4, while "" and "localhost" are the address-family-agnostic
equivalents.  Switch PuTTY to using the latter, since it doesn't seem
right to force IPv4.

There's an argument that PuTTY should provide a means of configuring the
address family used for remote forwardings like it does for local ones.

[originally from svn r9668]

11 years agoTwo related changes to timing code:
Ben Harris [Tue, 18 Sep 2012 21:42:48 +0000 (21:42 +0000)]
Two related changes to timing code:

First, make absolute times unsigned.  This means that it's safe to
depend on their overflow behaviour (which is undefined for signed
integers).  This requires a little extra care in handling comparisons,
but I think I've correctly adjusted them all.

Second, functions registered with schedule_timer() are guaranteed to be
called with precisely the time that was returned by schedule_timer().
Thus, it's only necessary to check these values for equality rather than
doing risky range checks, so do that.

The timing code still does lots that's undefined, unnecessary, or just
wrong, but this is a good start.

[originally from svn r9667]

11 years agoQuell a (correct) GCC warning in the NO_HTMLHELP case.
Ben Harris [Mon, 17 Sep 2012 22:28:07 +0000 (22:28 +0000)]
Quell a (correct) GCC warning in the NO_HTMLHELP case.

[originally from svn r9666]

11 years agoShould have been part of r9663: do use the platform-independent version
Ben Harris [Thu, 13 Sep 2012 23:00:29 +0000 (23:00 +0000)]
Should have been part of r9663: do use the platform-independent version
of smemclr when compiling with Winelib.

[originally from svn r9665]
[r9663 == 3b27c3e32b8fe560c5b2bb35781038715abf29c2]

11 years agoTweak comment in Recipe that had become separated from its code.
Ben Harris [Thu, 13 Sep 2012 22:34:53 +0000 (22:34 +0000)]
Tweak comment in Recipe that had become separated from its code.

[originally from svn r9664]

11 years agoDon't try to use SecureZeroMemory under Winelib, since it isn't available
Ben Harris [Thu, 13 Sep 2012 22:34:18 +0000 (22:34 +0000)]
Don't try to use SecureZeroMemory under Winelib, since it isn't available
there (yet).

[originally from svn r9663]

11 years agoDefine SECURITY_WIN32 for Winelib/Cygwin builds as well as for VC.
Ben Harris [Thu, 13 Sep 2012 22:33:27 +0000 (22:33 +0000)]
Define SECURITY_WIN32 for Winelib/Cygwin builds as well as for VC.

This should perhaps go into winmisc.c: it's caused problems for
other people too:

<http://stackoverflow.com/questions/8530159/vs2010-build-error-at-putty-source>

[originally from svn r9662]

11 years agoIn the cygwin Makefile, use "-o" on the windres command line to specify
Ben Harris [Thu, 13 Sep 2012 21:53:47 +0000 (21:53 +0000)]
In the cygwin Makefile, use "-o" on the windres command line to specify
the output file.  This appears to be supported by windres, and is
required by wrc (the Winelib resource compiler).

[originally from svn r9661]

11 years agoMost of the code for "nc" mode duplicated that for opening a session or
Ben Harris [Sat, 8 Sep 2012 19:46:07 +0000 (19:46 +0000)]
Most of the code for "nc" mode duplicated that for opening a session or
a fowarded port.  Arrange that this code is shared instead.  The main
visible change is a slight change of log messages.

[originally from svn r9655]

11 years agoHiroshi Oota points out that PuTTY's agent forwarding sockets can get
Simon Tatham [Sat, 8 Sep 2012 10:40:36 +0000 (10:40 +0000)]
Hiroshi Oota points out that PuTTY's agent forwarding sockets can get
confused if they receive a request followed by immediate EOF, since we
currently send outgoing EOF as soon as we see the incoming one - and
then, when the response comes back from the real SSH agent, we send it
along anyway as channel data in spite of having sent EOF.

To fix this, I introduce a new field for each agent channel which
counts the number of calls to ssh_agentf_callback that are currently
expected, and we don't send EOF on an agent channel until we've both
received EOF and that value drops to zero.

[originally from svn r9651]

11 years agoFactor out some common code for constructing SSH2_CHANNEL_OPEN.
Ben Harris [Sat, 1 Sep 2012 12:28:38 +0000 (12:28 +0000)]
Factor out some common code for constructing SSH2_CHANNEL_OPEN.

[originally from svn r9649]

11 years agostruct winadj is unused now. G/c it.
Ben Harris [Sat, 1 Sep 2012 12:10:27 +0000 (12:10 +0000)]
struct winadj is unused now.  G/c it.

[originally from svn r9648]

11 years agoAll of the initial CHANNEL_REQUESTs are conditional on (ssh->mainchan &&
Ben Harris [Sat, 1 Sep 2012 12:03:12 +0000 (12:03 +0000)]
All of the initial CHANNEL_REQUESTs are conditional on (ssh->mainchan &&
!ssh->ncmode), so bundle them up in a big block conditional on this rather
than checking it five times.

[originally from svn r9647]

11 years agodo_ssh2_authconn() now installs the standard handlers for CHANNEL_SUCCESS
Ben Harris [Sat, 1 Sep 2012 11:32:53 +0000 (11:32 +0000)]
do_ssh2_authconn() now installs the standard handlers for CHANNEL_SUCCESS
and CHANNEL_FAILURE as soon as it's opened a channel, so there's no need
for it to set them again later.

[originally from svn r9646]

11 years agoMemory leak fixes reported by Balazs Domjan.
Simon Tatham [Thu, 30 Aug 2012 18:48:08 +0000 (18:48 +0000)]
Memory leak fixes reported by Balazs Domjan.

[originally from svn r9645]

11 years agoRename the various ssh2_maybe_setup_* functions to ssh2_setup_*, and
Simon Tatham [Thu, 30 Aug 2012 18:44:35 +0000 (18:44 +0000)]
Rename the various ssh2_maybe_setup_* functions to ssh2_setup_*, and
move the primary conditions out of them into their callers. Fixes a
crash in 'plink -N', since those functions would be called with a NULL
channel parameter and immediately dereference it to try to get c->ssh.

[originally from svn r9644]

11 years agoIn openssh_read(), we shouldn't ever return SSH2_WRONG_PASSPHRASE for
Simon Tatham [Thu, 30 Aug 2012 18:44:34 +0000 (18:44 +0000)]
In openssh_read(), we shouldn't ever return SSH2_WRONG_PASSPHRASE for
an unencrypted key. (The other import function, sshcom_read(), already
got this right.) Thanks to David Wedderwille for the report.

This is more than just an error-reporting mistake; it actually causes
Windows PuTTYgen to tight-loop on attempting to load a corrupt OpenSSH
key, because the 'wrong passphrase' return value causes the caller to
loop round and try again, but of course it knows the key is
unencrypted so it doesn't prompt for a different passphrase and just
tries again with no change...

[originally from svn r9643]

11 years agoAvoid leaking file handles in load_openssh_key(), as reported by David
Simon Tatham [Thu, 30 Aug 2012 18:44:33 +0000 (18:44 +0000)]
Avoid leaking file handles in load_openssh_key(), as reported by David
Wedderwille.

[originally from svn r9642]

11 years agoFix a controlling-terminal bug reported by Anthony Heading: Cygwin
Simon Tatham [Tue, 28 Aug 2012 17:42:47 +0000 (17:42 +0000)]
Fix a controlling-terminal bug reported by Anthony Heading: Cygwin
doesn't have TIOCSCTTY, so my attempt to set the ctty of the child
process isn't doing anything, and only works by chance when you run
bash because bash does the thing that _will_ set the ctty, namely
opening the terminal file again without O_NOCTTY. So now we do that
too.

[originally from svn r9638]

11 years agoPartially revert r9636. It is true that we can directly return the
Simon Tatham [Tue, 28 Aug 2012 17:41:10 +0000 (17:41 +0000)]
Partially revert r9636. It is true that we can directly return the
result of memcmp, but untrue that we can do so _unconditionally_: if
memcmp returns zero, we still need to fall through to the next
comparison.

[originally from svn r9637]
[r9636 == 538090ede444b111039904cf918f095dc858e875]

11 years agoClang, like LCC, objects to using '<' and '>' on function pointers.
Ben Harris [Mon, 27 Aug 2012 23:16:49 +0000 (23:16 +0000)]
Clang, like LCC, objects to using '<' and '>' on function pointers.

I'm not entirely sure that using memcmp() is any more defined by the C
standard, but at least Clang doesn't complain about it.  While I'm
here, tree234 doesn't require that comparison functions return
precisely +1 or -1, so we can use the return value of memcmp()
directly.

[originally from svn r9636]

11 years agoChange return type of do_ssh2_transport() to void.
Ben Harris [Mon, 27 Aug 2012 22:02:17 +0000 (22:02 +0000)]
Change return type of do_ssh2_transport() to void.

Nothing pays attention to it any more, anyway.

[originally from svn r9635]

11 years agoMake bombout() less of a macro and more of a function.
Ben Harris [Mon, 27 Aug 2012 21:55:45 +0000 (21:55 +0000)]
Make bombout() less of a macro and more of a function.

This gives GCC slightly fewer opportunities to gratuitously inflate
its output.

[originally from svn r9634]

11 years agoFix a memory leak in parse_ttymodes() (found by Memcheck/Valgrind).
Ben Harris [Mon, 27 Aug 2012 19:11:39 +0000 (19:11 +0000)]
Fix a memory leak in parse_ttymodes() (found by Memcheck/Valgrind).

[originally from svn r9633]

11 years agoSlightly simplify crFinishFree{,V}.
Ben Harris [Mon, 27 Aug 2012 18:44:06 +0000 (18:44 +0000)]
Slightly simplify crFinishFree{,V}.

They're only likely to be useful for freeing a coroutine state
structure, in which case there's no need to reset the line number
(since all such coroutines keep their line number in the state
structure) and the state structure pointer is always called "s".

[originally from svn r9632]

11 years agoDon't close SSH-2 channels with outstanding channel requests on local error.
Ben Harris [Mon, 27 Aug 2012 17:37:44 +0000 (17:37 +0000)]
Don't close SSH-2 channels with outstanding channel requests on local error.

In sshfwd_unclean_close(), get ssh2_check_close() to handle sending
SSH_MSG_CHANNEL_CLOSE.  That way, it can hold off doing so until any
outstanding channel requests are processed.

Also add event log message for unclean channel closures.

[originally from svn r9631]

11 years agoIt's not legal to free a coroutine's state structure before invoking
Simon Tatham [Mon, 27 Aug 2012 14:34:41 +0000 (14:34 +0000)]
It's not legal to free a coroutine's state structure before invoking
crFinish or crFinishV, since they will attempt to write to the
coroutine state variable contained in that structure. Introduced some
new all-in-one macros crFinishFree and crFinishFreeV, and used those
instead. Should fix today's report of a crash just after authentication.

[originally from svn r9630]

11 years agoFix markup error introduced in r9626.
Jacob Nevins [Sun, 26 Aug 2012 09:50:57 +0000 (09:50 +0000)]
Fix markup error introduced in r9626.

[originally from svn r9627]
[r9626 == 92fc25c6cd04788cd409021979106f98ac032f30]

11 years agoRemove documentation for "Out of space for port forwardings" error.
Ben Harris [Sun, 26 Aug 2012 09:19:34 +0000 (09:19 +0000)]
Remove documentation for "Out of space for port forwardings" error.

It no longer exists in the code, and should have been obsoleted by
r9214 at the latest.

[originally from svn r9626]
[r9214 == a1f3b7a358adaa7c2a98359cd0373aa823eeb14b]

11 years agoImprove window-size handling in Unix Plink.
Ben Harris [Sat, 25 Aug 2012 22:57:39 +0000 (22:57 +0000)]
Improve window-size handling in Unix Plink.

Unconditionally override the configured terminal size with the one
from stdin if it's available.  This avoids the silliness whereby if
Default Settings had a terminal size set, Plink used this and thus
caused the server to use the wrong size.

[originally from svn r9624]

11 years agoBetter handling of outstanding CHANNEL_REQUESTS on channel destruction.
Ben Harris [Sat, 25 Aug 2012 21:06:48 +0000 (21:06 +0000)]
Better handling of outstanding CHANNEL_REQUESTS on channel destruction.

Part the first: make sure that all structures describing channel
requests are freed when the SSH connection is freed.  This involves
adding a means to ask a response handler to free any memory it holds.

Part the second: in ssh_channel_try_eof(), call
ssh2_channel_check_close() rather than emitting an SSH_MSG_CHANNEL_EOF
directly.  This avoids the possibility of closing the channel while a
CHANNEL_REQUEST is outstanding.

Also add some assertions that helped with tracking down the latter
problem.

[originally from svn r9623]

11 years agoFactor out common code to construct CHANNEL_REQUESTS.
Ben Harris [Sat, 25 Aug 2012 15:57:05 +0000 (15:57 +0000)]
Factor out common code to construct CHANNEL_REQUESTS.

This reduces code size a little and also makes it harder to
accidentally request a reply without putting in place a handler for
it or vice versa.

[originally from svn r9620]

11 years agoSimplify handling of responses to channel requests.
Ben Harris [Sat, 25 Aug 2012 15:04:29 +0000 (15:04 +0000)]
Simplify handling of responses to channel requests.

The various setup routines can only receive CHANNEL_SUCCESS or
CHANNEL_FAILURE, so there's no need for the to worry about receiving
anything else.  Strange packets will end up in do_ssh2_authconn
instead.

[originally from svn r9619]

11 years agoAdd some kind of window tracking to logparse.pl.
Ben Harris [Sat, 25 Aug 2012 14:34:20 +0000 (14:34 +0000)]
Add some kind of window tracking to logparse.pl.

[originally from svn r9618]

11 years agoFix matching of channel and global requests with replies in logparse.pl.
Ben Harris [Sat, 25 Aug 2012 13:54:56 +0000 (13:54 +0000)]
Fix matching of channel and global requests with replies in logparse.pl.

In each case, want_reply was being treated as true even when it wasn't,
because it got decoded into "yes"/"no", both of which are true in
Perl.

[originally from svn r9617]

11 years agoHandle all replies to CHANNEL_REQUESTs through the per-channel queue.
Ben Harris [Sat, 25 Aug 2012 13:39:32 +0000 (13:39 +0000)]
Handle all replies to CHANNEL_REQUESTs through the per-channel queue.

Each of the minor start-of-session requests is now dealt with by its own
little co-routine, while the shell/command is done in do_ssh2_authconn()
itself.  This eliminates one more round-trip in session setup: PuTTY gets
all the way up to sending a shell request before worrying about any
replies.

[originally from svn r9616]

11 years agoGeneralise SSH_MSG_CHANNEL_{SUCCESS,FAILURE} handling.
Ben Harris [Sat, 25 Aug 2012 11:12:14 +0000 (11:12 +0000)]
Generalise SSH_MSG_CHANNEL_{SUCCESS,FAILURE} handling.

Now each channel has a queue of arbitrary handlers for those messages,
with anything that sends a CHANNEL_REQUEST with want_reply true pushing
a new entry onto the queue, and a shared handler that dispatches
responses appropriately.

Currently, this is only used for winadj@putty.projects.tartarus.org, but
extending it to cover the initial requests as well shouldn't be too
painful.

[originally from svn r9615]

11 years agoFix a bug in the PSFTP command parser which would cause it to
Simon Tatham [Wed, 22 Aug 2012 18:36:44 +0000 (18:36 +0000)]
Fix a bug in the PSFTP command parser which would cause it to
hallucinate an extra empty argument word at the end of a line if the
line ended in more than one whitespace character.

[originally from svn r9613]

11 years agoEnhance logparse.pl so that it detects channels running SFTP and
Simon Tatham [Wed, 22 Aug 2012 18:25:28 +0000 (18:25 +0000)]
Enhance logparse.pl so that it detects channels running SFTP and
decodes the SFTP message layer in addition to the underlying SSH.
Requests and responses are matched up via their ids.

[originally from svn r9612]

11 years agoFix recently-introduced memory leak in ssh2_msg_unexpected().
Ben Harris [Tue, 21 Aug 2012 23:04:22 +0000 (23:04 +0000)]
Fix recently-introduced memory leak in ssh2_msg_unexpected().

[originally from svn r9611]

11 years agoDe-duplicate code in KEXINIT generation.
Ben Harris [Tue, 21 Aug 2012 22:33:31 +0000 (22:33 +0000)]
De-duplicate code in KEXINIT generation.

There's no need to have identical code generating server-to-client and
client-to-server versions of the cipher and MAC lists; a couple of
twice-around loops will do fine.

[originally from svn r9610]

11 years agoReworking of packet delivery to coroutines in SSH-2.
Ben Harris [Tue, 21 Aug 2012 22:04:56 +0000 (22:04 +0000)]
Reworking of packet delivery to coroutines in SSH-2.

Before, NULL in the dispatch table meant "send to the appropriate one of
do_ssh2_transport() and do_ssh2_authconn()".  Now those (via small
shims) are specified directly in the dispatch table, so ssh2_protocol()
is much simpler.

In the process, this has somewhat centralised the handling of gross
server protocol violations.  PuTTY will now disconnect with a rude
message when (e.g.) OpenSSH sends us an SSH_MSG_UNIMPLEMENTED when we
try to KEXINIT during authentication.

[originally from svn r9609]

11 years agoFix an erroneous "case" fallthrough in ssh1_msg_channel_close, which was
Ben Harris [Sun, 19 Aug 2012 11:35:26 +0000 (11:35 +0000)]
Fix an erroneous "case" fallthrough in ssh1_msg_channel_close, which was
causing assertion failures when closing X11 channels in SSH-1.  Also fix
another pasto.

[originally from svn r9608]

11 years agoTweak to SSH coroutine code: put line number in the coroutine state
Ben Harris [Sat, 18 Aug 2012 09:30:01 +0000 (09:30 +0000)]
Tweak to SSH coroutine code: put line number in the coroutine state
structure, which is consistent with Simon's canonical version of the
macros.

[originally from svn r9607]

11 years agoUse a single sftp_senddata() to send each SFTP packet, rather than
Simon Tatham [Sun, 12 Aug 2012 20:17:13 +0000 (20:17 +0000)]
Use a single sftp_senddata() to send each SFTP packet, rather than
using one for the length field and one for the rest of the packet
contents. Since sftp_senddata() has no queuing or deferral mechanism
but instead constructs and sends an SSH2_MSG_CHANNEL_DATA message
immediately, this change has the effect of ceasing to split every SFTP
packet across two SSH messages.

[originally from svn r9603]

11 years agoRework bufchain code to allow for variable-sized granules.
Ben Harris [Sat, 11 Aug 2012 09:10:31 +0000 (09:10 +0000)]
Rework bufchain code to allow for variable-sized granules.
bufchain_add() now allocates at most one new granule.  Granules still
have a minimum size, so small adds still get coalesced.

The main practical consequence of this is that PSCP and PSFTP now
generate 4K SSH packets rather than 512-byte ones.  Also, the compiled
code (on my Ubuntu box) is fractionally smaller.

[originally from svn r9602]

11 years agoReduce the number of round-trips involved in opening an SSH-2 session
Ben Harris [Thu, 2 Aug 2012 22:18:18 +0000 (22:18 +0000)]
Reduce the number of round-trips involved in opening an SSH-2 session
by sending most of the initial SSH_MSG_CHANNEL_REQUEST messages before
waiting for any replies.  The initial version of this code was a clever
thing with a two-pass loop, but that got hairy so I went for the simpler
approach of separating the request and reply code and having flags to
keep track of which requests have been sent.

[originally from svn r9599]

11 years agoAdd a missing \define I accidentally assumed was there in r9592.
Simon Tatham [Sun, 29 Jul 2012 13:38:31 +0000 (13:38 +0000)]
Add a missing \define I accidentally assumed was there in r9592.

[originally from svn r9593]
[r9592 == 71f0c7546dfa8930277c895b88f57585fd8ef854]

11 years agoAdd a bug-compatibility flag to disable the
Simon Tatham [Sat, 28 Jul 2012 19:30:12 +0000 (19:30 +0000)]
Add a bug-compatibility flag to disable the
winadj@putty.projects.tartarus.org request. Not currently enabled
automatically, but should be usable as a manual workaround.

[originally from svn r9592]

11 years agoFix an embarrassing mistake in config box handling which was causing
Simon Tatham [Sat, 28 Jul 2012 16:53:09 +0000 (16:53 +0000)]
Fix an embarrassing mistake in config box handling which was causing
changes to any SSH bug config option to be lost when the config box
switched to a different panel, at least on GTK.

[originally from svn r9591]

11 years agoMove the declaration of smemclr() out of putty.h into misc.h, because
Simon Tatham [Sat, 28 Jul 2012 16:33:51 +0000 (16:33 +0000)]
Move the declaration of smemclr() out of putty.h into misc.h, because
one of its uses (in sshaes.c) wasn't picking up the former. Thanks to
Ubuntu's gcc for spotting that.

[originally from svn r9590]

11 years agoRemove an unused variable.
Simon Tatham [Sat, 28 Jul 2012 16:33:17 +0000 (16:33 +0000)]
Remove an unused variable.

[originally from svn r9589]

11 years agoIntroduce a new utility function smemclr(), which memsets things to
Simon Tatham [Sun, 22 Jul 2012 19:51:50 +0000 (19:51 +0000)]
Introduce a new utility function smemclr(), which memsets things to
zero but does it in such a way that over-clever compilers hopefully
won't helpfully optimise the call away if you do it just before
freeing something or letting it go out of scope. Use this for
(hopefully) every memset whose job is to destroy sensitive data that
might otherwise be left lying around in the process's memory.

[originally from svn r9586]

11 years agoRemove a bashism in mksrcarc.sh, without which bob builds fail on
Simon Tatham [Sun, 22 Jul 2012 19:48:39 +0000 (19:48 +0000)]
Remove a bashism in mksrcarc.sh, without which bob builds fail on
Ubuntu (whose /bin/sh is not bash).

[originally from svn r9585]

11 years agoAdd '-Wall -Werror' to the compile options in the autotools makefile,
Simon Tatham [Thu, 19 Jul 2012 04:42:34 +0000 (04:42 +0000)]
Add '-Wall -Werror' to the compile options in the autotools makefile,
having just noticed that Makefile.gtk had it and this one doesn't. (Of
course, this being autoconf, we can easily enough make it conditional
on the compiler actually being gcc.)

[originally from svn r9583]

11 years agoRework the new type-check in sresize so that it doesn't cause a
Simon Tatham [Thu, 19 Jul 2012 04:29:50 +0000 (04:29 +0000)]
Rework the new type-check in sresize so that it doesn't cause a
compile warning ('left-hand operand of comma expression has no
effect'), which of course becomes fatal under -Werror.

(This would have been instantly noticeable to people compiling with
the old-fashioned Makefile.gtk, which does include -Wall -Werror, but
those of us using the new autoconf makefile hadn't noticed.)

[originally from svn r9582]

11 years agoRemove term_key, which was intended to handle function keys in a cross-
Owen Dunn [Wed, 18 Jul 2012 16:08:54 +0000 (16:08 +0000)]
Remove term_key, which was intended to handle function keys in a cross-
platform manner, but which nothing ever called.  It thus served only to
trap up the unwary.  The live function key handling code lives in the
frontends, i.e. window.c on Windows and gtkwin.c on Unix.

[originally from svn r9579]

11 years agoIf pterm's execvp fails when given the whole argument list after -e,
Simon Tatham [Wed, 11 Jul 2012 18:12:17 +0000 (18:12 +0000)]
If pterm's execvp fails when given the whole argument list after -e,
and the argument list contains only one string, try again by passing
that single string to "$SHELL -c" to be parsed as a shell command.
This matches xterm's behaviour (as of xterm 261, at least), and means
in practice that users can do _either_ of 'pterm -e some command' and
'pterm -e "some command"'.

(A quick survey suggests that the majority of X terminal programs agree
with pterm's old behaviour of only supporting '-e some command',
except that gnome-terminal only supports the other behaviour and xterm
supports both. With that disagreement, I think supporting both is
probably the sensible thing.)

[originally from svn r9575]

11 years agoTurns out that the compose-keys fix in r9567 did in fact break one
Simon Tatham [Thu, 5 Jul 2012 23:45:20 +0000 (23:45 +0000)]
Turns out that the compose-keys fix in r9567 did in fact break one
piece of keyboard handling: if Num Lock is on, numeric keypad keys are
eaten by the IM, so we must avoid passing them to the IM in the first
place if we're in any non-default numeric keypad mode (application or
Nethack).

This is a grubby way to do it, but the more obvious approach of just
moving the Nethack and app-keypad if statements up to above the IM
call doesn't work because those statements depend on the generic
Alt-prefix handling that happens just _below_ the IM call. So instead
I just repeat the list of keystrokes and modes in an if statement
conditionalising the IM call.

[originally from svn r9573]
[r9567 == 7fc8db15b262ba644b1c7f8e9b89cc5a86f104f2]

11 years agoStop including <gtk/gtkcontainer.h> directly, since a user points out
Simon Tatham [Sun, 1 Jul 2012 16:55:36 +0000 (16:55 +0000)]
Stop including <gtk/gtkcontainer.h> directly, since a user points out
that it'll be absent in GTK 3.

[originally from svn r9571]

11 years agoJohn Hartnup reports that Apache SSHD's SFTP server responds to
Simon Tatham [Wed, 20 Jun 2012 17:39:32 +0000 (17:39 +0000)]
John Hartnup reports that Apache SSHD's SFTP server responds to
FXP_READDIR on an empty directory by returning a zero-length list of
filenames, instead of the more common response of a list containing
just "." and "..". Stop PSFTP failing an assertion when that happens.

[originally from svn r9569]

11 years agoAdd the missing code to treat data coming from the input method as
Simon Tatham [Mon, 18 Jun 2012 18:10:59 +0000 (18:10 +0000)]
Add the missing code to treat data coming from the input method as
keypresses for purposes of hiding the mouse pointer and resetting the
scrollback.

[originally from svn r9568]

11 years agoSupport for dead keys and compose sequences on Unix, by instantiating
Simon Tatham [Sun, 17 Jun 2012 07:26:23 +0000 (07:26 +0000)]
Support for dead keys and compose sequences on Unix, by instantiating
a GtkIMMulticontext and having that filter most keypresses. I think
I've got this right so that it doesn't break any previous deliberate
keyboard-handling behaviour that's now _after_ the 'if (filter
keypress) return' statement.

[originally from svn r9567]

11 years agoIntroduce a third setting for the 'bold as colour' mode, which lets
Simon Tatham [Sat, 9 Jun 2012 15:09:22 +0000 (15:09 +0000)]
Introduce a third setting for the 'bold as colour' mode, which lets
you both brighten the colour _and_ bold the font at the same time.
(Fixes 'bold-font-colour' and Debian #193352.)

[originally from svn r9559]

12 years agoAdd missing check for failure to agree a host key algorithm.
Simon Tatham [Mon, 4 Jun 2012 23:32:36 +0000 (23:32 +0000)]
Add missing check for failure to agree a host key algorithm.

[originally from svn r9557]

12 years agoMissing #include.
Simon Tatham [Fri, 1 Jun 2012 19:43:05 +0000 (19:43 +0000)]
Missing #include.

[originally from svn r9551]

12 years agoFix a bug in cygtermd, spotted by Casey Zacek, in which we
Simon Tatham [Fri, 1 Jun 2012 06:50:37 +0000 (06:50 +0000)]
Fix a bug in cygtermd, spotted by Casey Zacek, in which we
unconditionally set the telnet state to SEENCR regardless of whether
we have actually seen a CR, and as a result sending a NUL through
PuTTY (via Ctrl-Space or whatever) does not work. Must have arisen
through some kind of really weird cut-and-paste error!

[originally from svn r9545]

12 years agoEnhance my 'sresize' macro so that it type-checks the pointer you pass
Simon Tatham [Fri, 18 May 2012 19:51:11 +0000 (19:51 +0000)]
Enhance my 'sresize' macro so that it type-checks the pointer you pass
_in_ to it, as well as the one it returns. Why have I never thought of
doing that before?!

[originally from svn r9539]

12 years agoFix bug in the new CLOCK_MONOTONIC implementation. I was treating the
Simon Tatham [Tue, 15 May 2012 22:19:21 +0000 (22:19 +0000)]
Fix bug in the new CLOCK_MONOTONIC implementation. I was treating the
nanoseconds field as a microseconds field, with hilarious consequences.

[originally from svn r9535]