Simon Tatham [Fri, 4 Apr 2008 10:56:26 +0000 (10:56 +0000)]
Reinstate all the GTK1-specific code under ifdefs, and verify that
we can now build and run successfully using both GTK1 and GTK2 by
giving appropriate options to make. (Specifically, to override the
default of GTK2 in favour of GTK1, "make GTK_CONFIG=gtk-config".)
Simon Tatham [Fri, 4 Apr 2008 10:16:24 +0000 (10:16 +0000)]
Rename a structure field to avoid clashing with one of the old GTK1
ones. (I'm going to merge the GTK1 list code back in under ifdefs,
and I want none of the disputed structure fields to have the same
names, so that I'll reliably be told by the compiler if I keep the
wrong piece of code outside the ifdef.)
Simon Tatham [Wed, 2 Apr 2008 17:32:17 +0000 (17:32 +0000)]
Update autoconf for GTK 2. We now check for both GTK2 and GTK1, and
in the presence of GTK 2 we also check to see whether we have a
prehistoric Pango (since Pango itself helpfully doesn't provide that
functionality, bah).
Simon Tatham [Wed, 2 Apr 2008 17:04:21 +0000 (17:04 +0000)]
Another tedious chore off the to-do list. I've just checked over my
custom Columns layout class to see what fiddly details of
GTK2isation were yet to be done. It turns out that all the basic
object management got moved out of GTK into a separate library, so
that all the gtk_object_* calls are deprecated and g_object_* should
be used instead; having done that, though, it all looks perfectly
fine.
Simon Tatham [Wed, 2 Apr 2008 16:26:01 +0000 (16:26 +0000)]
TODO update: I don't think it's worth switching to GTK2's native
shortcut mechanism. The existing code doesn't use any deprecated
calls, and translating shortcut text _into_ Pango markup just sounds
too unpleasant to do if I don't actually have to. Not to mention
that the documentation for the Pango markup language doesn't tell me
how to distinguish a mnemonic underscore prefix from a literal
underscore in label text, but I know my current code can get that
right (the current config box talks about TCP_NODELAY and
SO_KEEPALIVE in widget labels that also have functioning shortcuts).
Simon Tatham [Wed, 2 Apr 2008 14:50:47 +0000 (14:50 +0000)]
Enable the display of server-side font aliases by default in my font
selector. I had previously been worried that the default of not
showing aliases interacted badly with the default actual font
_being_ specified as an alias. One of those defaults had to change,
and I've decided which: `fixed' is staying as Unix PuTTY's default
font in defiance of GTK2's vigorous encouragement of Pango.
Simon Tatham [Wed, 2 Apr 2008 14:48:06 +0000 (14:48 +0000)]
Update all the list box code in gtkdlg.c to use the new-style GTK2
GtkTreeView, GtkComboBox and GtkComboBoxEntry instead of the various
old deprecated stuff. Immediate benefit: GTK2 natively supports real
drag lists, hooray!
Simon Tatham [Sat, 29 Mar 2008 20:02:12 +0000 (20:02 +0000)]
I give up. I can't work out what the purpose of the call to
gtk_container_dequeue_resize_handler in request_resize() was;
everything seems to work fine without it. So I'm removing the
nonportable GTK 2 instance of it, and if anything ever goes wrong as
a result then I'll at least find out what the problem was.
Simon Tatham [Sat, 29 Mar 2008 15:44:32 +0000 (15:44 +0000)]
Be more picky than Pango when validating a Pango font description
string. Without this, Richard B reports that Pango 1.18 will treat
_anything_ as valid, which means PuTTY can never fall back to X
fonts.
Simon Tatham [Sat, 29 Mar 2008 14:54:55 +0000 (14:54 +0000)]
Aha, _that's_ why there was some unexplained space on the RHS of the
font selector: I had got the row and column counts in
gtk_table_new() back to front, so the space on the right was the
padding around five empty table columns! (And apparently a GtkTable
silently expands if you try to use rows that don't exist, which is
why I hadn't already noticed.)
Fixed that, and added some padding around the entire table. I think
my font selector is now finished, except for any bug fixes that come
up in testing.
Simon Tatham [Sat, 29 Mar 2008 14:21:25 +0000 (14:21 +0000)]
Deal with the possibility of no valid font being selected at all
during an entire run of unifontsel (because unifontsel_set_name was
either not called at all, or called with a name that didn't
correspond to any known font). In this situation we grey out the OK
button until a valid font is selected, and we have
unifontsel_get_name return NULL rather than failing an assertion if
it should be called in that state. The current client code in
gtkdlg.c should never encounter a NULL return, since it only calls
it after the OK button is clicked, but I've stuck an assertion in
there too on general principles.
Simon Tatham [Sat, 29 Mar 2008 10:48:16 +0000 (10:48 +0000)]
Detect non-monospaced X fonts, and respond by drawing text one
character at a time centred in its character cell, as we do for
Pango. Gives much better results for those non-monospaced fonts
which are usable as terminal fonts, and shows up the problems with
the others more readily. (In particular, this means the preview pane
in the font selector now warns you there will be trouble if you
select such a font.)
Simon Tatham [Sat, 29 Mar 2008 10:16:48 +0000 (10:16 +0000)]
When the user switches between fonts using the font family or style
selectors, preserve their most recent size selection as faithfully
as possible. We do this by having a secondary size variable
indicating what they _intend_, so we can come back to their intended
size even after going through a font which doesn't include it.
Simon Tatham [Thu, 27 Mar 2008 19:53:28 +0000 (19:53 +0000)]
Move the font-preview updating code out into a separate function so
we can call it both when the drawing area changes size and when the
selected font changes. As a result, the preview pane doesn't start
off blank any more.
Simon Tatham [Thu, 27 Mar 2008 19:41:08 +0000 (19:41 +0000)]
More consistent handling of X11 font aliases: we now don't resolve
them automatically. If the user selects an alias in the font
selector, they get that alias copied literally into the output font
name string; when they return to the font selector, the alias is
still selected. We still _can_ resolve aliases, but we only do it on
demand: double-clicking one in the list box will do the job.
Simon Tatham [Wed, 26 Mar 2008 20:20:25 +0000 (20:20 +0000)]
Sort the styles of Pango font families into a sensible order,
instead of alphabetical order. This is more than cosmetic: it's
important because the first one in the list is selected by default.
Simon Tatham [Wed, 26 Mar 2008 18:30:20 +0000 (18:30 +0000)]
Add ifdefs for older versions of GTK2 and Pango. Unfortunately, the
latter require manual input to the Makefile, since the Pango
developers in their unbounded wisdom (that is, unbounded below)
didn't bother to start providing the PANGO_VERSION macros until
release 1.16 - ten releases _after_ everything I'm trying to check!
Simon Tatham [Tue, 25 Mar 2008 21:49:14 +0000 (21:49 +0000)]
Unified font selector dialog box. _Extremely_ unfinished - there's a
sizable TODO at the top of gtkfont.c - but it's basically functional
enough to select fonts of both types, so I'm checking it in now
before I accidentally break it.
Simon Tatham [Sat, 22 Mar 2008 18:11:17 +0000 (18:11 +0000)]
Implemented a Pango back end. GTK 2 PuTTY can now switch seamlessly
back and forth between X fonts and Pango fonts, provided you're
willing to type in the names of the former by hand.
Simon Tatham [Sat, 22 Mar 2008 12:01:16 +0000 (12:01 +0000)]
For convenience of debugging, and perhaps some real convenience at
some point too: introduce a bunch of environment variables which can
override Unix PuTTY's usual idea of where to find its dotfiles.
Setting PUTTYDIR moves the entire ~/.putty directory; setting
PUTTYSESSIONS, PUTTYSSHHOSTKEYS or PUTTYRANDOMSEED move specific
things within that directory.
While I'm here, also be prepared to fall back to password file
lookups if $HOME is undefined (though we still use $HOME in
preference when it is defined, because that's polite and useful).
Also, on general principles, tweak the make_filename() function
prototype so it doesn't rely on fixed-size buffers.
Simon Tatham [Sat, 22 Mar 2008 11:40:23 +0000 (11:40 +0000)]
Refactor the font handling code: I've moved all the code that
explicitly deals with GdkFont out into a new module, behind a
polymorphic interface (done by ad-hoc explicit vtable management in
C). This should allow me to drop in a Pango font handling module in
parallel with the existing one, meaning that GTK2 PuTTY will be able
to seamlessly switch between X11 server-side fonts and Pango client-
side ones as the user chooses, or even use a mixture of the two
(e.g. an X11 font for narrow characters and a Pango one for wide
characters, or vice versa).
In the process, incidentally, I got to the bottom of the `weird bug'
mentioned in the old do_text_internal(). It's not a bug in
gdk_draw_text_wc() as I had thought: it's simply that GdkWChar is a
32-bit type rather than a 16-bit one, so no wonder you have to
specify twice the length to find all the characters in the string!
However, there _is_ a bug in GTK2's gdk_draw_text_wc(), which causes
it to strip off everything above the low byte of each GdkWChar,
sigh. Solution to both problems is to use an array of the underlying
Xlib type XChar2b instead, and pass it to gdk_draw_text() cast to
gchar *. Grotty, but it works. (And it'll become significantly less
grotty if and when we have to stop using the GDK font handling
wrappers in favour of going direct to Xlib.)
Simon Tatham [Fri, 7 Mar 2008 18:30:37 +0000 (18:30 +0000)]
Fix a cursor positioning infelicity.
The scenario: I start a small, say 80x24, pterm. I do some work in
it, generating plenty of scrollback, and eventually I `less' a file.
`less' switches to the alt screen. Then I want more vertical space
to look at the file, so I enlarge the window to more like 80x60.
When I quit `less' and switch back to the primary screen, some
scrollback has been pulled down into the screen, as expected - but
the saved _cursor position_ is still at line 24, not at the bottom
of the new terminal where the prompt it goes with has moved to.
Solution: term_size() should adjust the alt-screen saved cursor
positions as well as the normal cursor position.
(Curiously, the problem doesn't happen on my home Debian box, even
without this fix. It happens on my RH9 box at work, though.)
Simon Tatham [Thu, 21 Feb 2008 09:18:24 +0000 (09:18 +0000)]
Aha, _that's_ why I've been periodically getting blocking-write
problems using Unix PuTTY port forwarding. Sockets we create by
connect() are immediately set into nonblocking mode by fcntl, but
sockets we create by accept() were not. This trivial fix should help.
Ben Harris [Wed, 5 Dec 2007 00:02:06 +0000 (00:02 +0000)]
Add a new bug-compatibility mode that limits the window size we'll
advertise so that the server can't exceed our maximum packet size.
Enable it for "1.36_sshlib GlobalSCAPE" which apparently sends oversize
packets otherwise.
Simon Tatham [Mon, 26 Nov 2007 21:09:54 +0000 (21:09 +0000)]
sktree is indexed on the numeric value of the socket structure's
underlying WinSock SOCKET. Therefore, if we plan to modify the
SOCKET in a socket, we must remove it from the tree before doing so,
and put it back again afterwards. Otherwise it'll violate the tree's
sorting order, and sooner or later someone will try to find it and
get back NULL.
Jacob Nevins [Fri, 19 Oct 2007 21:47:47 +0000 (21:47 +0000)]
Marc TERRIER pointed out a couple of places that claim there is an X11
forwarding checkbox on the Tunnels panel, which hasn't been the case for
a while.
Ben Harris [Wed, 3 Oct 2007 20:29:27 +0000 (20:29 +0000)]
Take the code that does flow control in SSH-1, and make it work in SSH-2
as well. This won't be triggered in the usual case, but it's useful
if the remote end ignores our window, or if we're in "simple" mode and
setting the window far larger than is necessary.
Ben Harris [Tue, 2 Oct 2007 21:43:53 +0000 (21:43 +0000)]
More fixes to stdout and stderr. When the backlog on either clears, call
the backend's unthrottle function. If we don't, we'll deadlock. While
we're here, also pump as much data as possible out during each call to
try_output(), rather than restricting ourselves to a single call to
write().
Ben Harris [Tue, 2 Oct 2007 21:07:52 +0000 (21:07 +0000)]
As far as I can see (at least in NetBSD) O_NONBLOCK and FIONBIO are equivalent,
except that O_NONBLOCK is standardised and FIONBIO isn't. In consequence,
replace our only use of FIONBIO with O_NONBLOCK.
Inspired by Jonathan H N Chin, who had problems with this on Solaris.
Ben Harris [Mon, 1 Oct 2007 21:11:11 +0000 (21:11 +0000)]
Rather than rejecting spurious SSH_MSG_CHANNEL_SUCCESSes, and ignoring
spurious SSH_MSG_CHANNEL_FAILUREs, treat them as the protocol errors
they are and forcibly disconnect. Inspired by recent traffic on
comp.security.ssh.
Ben Harris [Sun, 30 Sep 2007 19:42:31 +0000 (19:42 +0000)]
Merge the looking up of channel numbers for SSH-2 channel messages into
a single function which also handles checking that channels exist and
are properly open. This should make PuTTY a little less tolerant of
servers that send bogus messages.
Ben Harris [Sun, 30 Sep 2007 12:45:49 +0000 (12:45 +0000)]
When writing session data to stdout or stderr, switch the relevant file
descriptor into non-blocking mode temporarily, and correctly handle returns
of EAGAIN from write(). This should fix unix-plink-stdout-nonblock, while
avoiding EAGAIN turning up where we aren't expecting it.
Ben Harris [Sat, 29 Sep 2007 12:27:45 +0000 (12:27 +0000)]
Add support for resetting the terminal modes on stderr to something sensible
before printing error messages to it. This should fix the stair-stepping
in Plink's progress messages.
Ben Harris [Mon, 24 Sep 2007 21:43:48 +0000 (21:43 +0000)]
My changes in r7738 (O_NONBLOCK for Unix Plink) were half-arsed, and
completely broke interactive logins. The problem, or at least one of the
problems, was that in interactive use stdin, stdout, and stderr tend to be
the same file, so setting O_NONBLOCK on the latter two also sets it on the
former. Thus, we need to cope with all of them being non-blocking.
Ben Harris [Mon, 24 Sep 2007 15:18:11 +0000 (15:18 +0000)]
Add support for automatically tuning the SSH-2 window size for decent
performance. The theory behind this is fairly simple, though the
implementation turns out to be a little trickier than it looks.
The basic idea is that when the connection isn't being limited by our ability
to process data, we want to ensure that the window size _as seen by the server_
never drops to zero. Measuring the server's view of the window size is done
by arranging for it to acknowledge every SSH_MSG_CHANNEL_WINDOW_ADJUST, or
rather an SSH_MSG_CHANNEL_REQUEST sent just before it. That way we can tell
when it its outgoing data stream it received the window adjustment, and
thus how small the server's view of the window got.
At present, we only ever increase the window size. In theory, we could
arrange to reduce it again if the server's view of it seemed to be persistently
too large, but my experiments suggest that getting this right will be tricky.
Ben Harris [Thu, 20 Sep 2007 21:33:21 +0000 (21:33 +0000)]
Don't try SSH-1 RSA authentication unless the server has advertised support
for it. It's possible that this obsoletes BUG_CHOKES_ON_RSA. Certainly
the one SSH-1.5-Cisco-1.25 server I found was correctly not advertising RSA
auth. For now, leave it in, because I'm not feeling entirely confident.
Ben Harris [Thu, 20 Sep 2007 21:07:24 +0000 (21:07 +0000)]
In SSH-1, don't attempt password authentication unless the server has
announced support for it. Instead exit with a fatal error (since password
auth is our last resort).
Ben Harris [Tue, 7 Aug 2007 22:02:03 +0000 (22:02 +0000)]
Correctly terminate nc target hostname when copying it.
While we're here, use memcpy rather than strncpy when we've already worked
out how much we're going to copy.
Ben Harris [Mon, 6 Aug 2007 20:56:52 +0000 (20:56 +0000)]
Use "int" rather than "unsigned" as the argument to ssh2_set_window, not
because it can ever be negative, but because we'll be comparing it with
another int. This way, C's promotion rules don't bite us and we should
stand slightly more chance of coping with broken servers that overrun our
window.
Ben Harris [Sun, 5 Aug 2007 14:18:43 +0000 (14:18 +0000)]
Small window-handling tweaks. Set the default big window to 0x7fffffff bytes,
and tweak ssh2_set_window() so it can cope with that. Also arrange to send
a private channel message in simple mode to tell the server that it can safely
use a large window too.
Ben Harris [Sat, 4 Aug 2007 22:14:19 +0000 (22:14 +0000)]
In the file-transfer applications, which only ever use the main
channel, arrange to set the SSH-2 window size to something very
large. This prevents the connection stalling when the window fills
up, and means that PSCP receives data _much_ faster.
Ben Harris [Sat, 4 Aug 2007 16:04:08 +0000 (16:04 +0000)]
Tweak window handling so that we send a window adjust if the window is half
used up, rather than over half. That this increases the throughput of PSCP
by 50% indicates just how broken our window handling is.
Ben Harris [Sat, 4 Aug 2007 15:48:52 +0000 (15:48 +0000)]
When omitting session data from logs, don't omit the length of the session
data string. This isn't strictly necessary, but it makes the logs easier to
use.
Ben Harris [Sat, 4 Aug 2007 14:32:06 +0000 (14:32 +0000)]
Tweak to window handling: Keep the local window in a signed integer, and
arrange to handle usefully the case where the server sends us more data
than it's allowed to. There's no danger of overflow, since the maximum is
OUR_V2_WINSIZE and the minimum is -OUR_V2_MAXPKT (at least if the server is
nice).
Ben Harris [Sun, 29 Jul 2007 14:02:00 +0000 (14:02 +0000)]
Don't throw away data that we receive before we're ready for it. Just save
it up for later. This should prevent hangs when talking to particularly
enthusiastic servers.
Ben Harris [Sat, 21 Jul 2007 21:39:36 +0000 (21:39 +0000)]
Split pkt_ctx into a separate enumeration for each of kex and userauth
instead of a bitfield for both. This doesn't gain much here, but it should
make it easier to make things other than logging use the context.
Ben Harris [Sat, 21 Jul 2007 13:43:57 +0000 (13:43 +0000)]
Arrange that log_packet() isn't called for raw data logging if logctx is null.
This allows us to send data in ssh_init(), albeit at the expense of its not
being properly logged, so arrange to send the version string then if that's
sensible, which should reduce the number of round-trips required to bring
up an SSH-2 connection.
Ben Harris [Thu, 19 Jul 2007 23:53:02 +0000 (23:53 +0000)]
Separate out the code for creating and sending SSH version strings so that in
the SSH-2-only case, we can send it as soon as we connect rather than waiting
for the server's one. Unfortunately, actually doing so will take a little
more effort -- there are subtleties to do with having a working log context
at the right moment that need to be sorted out.
Jacob Nevins [Sun, 1 Jul 2007 15:47:31 +0000 (15:47 +0000)]
Implement Marcin Bulandra's suggestion of only automatically updating the
port number in the GUI when the connection type is changed if the current
port number is the standard one for the current protocol.
It's not perfect, but it should make the common case of tabbing through the
Session panel easier when starting non-SSH connections on odd ports.
Jacob Nevins [Sun, 1 Jul 2007 15:41:09 +0000 (15:41 +0000)]
Remove port number validation from Windows PuTTY -- it could cause unnecessary
trouble with serial connections, and a port number of zero gets caught later
anyway.
Jacob Nevins [Sat, 30 Jun 2007 21:56:44 +0000 (21:56 +0000)]
Rationalise access to, and content of, backends[] array.
Should be no significant change in behaviour.
(Well, entering usernames containing commas on Plink's command line will be
a little harder now.)
Simon Tatham [Tue, 22 May 2007 18:37:17 +0000 (18:37 +0000)]
Retire the e-gold link. Nobody's used it in years; I honestly don't
know how I'd go about retrieving money from them any more because my
last exchange transaction went through a company who subsequently
turned out to be dodgy; and a user points out that e-gold is in
legal trouble, which suggests that avoiding it is probably wise.