KERMIT 95 FIXED BUG LIST
The Kermit Project, Columbia University
http://www.columbia.edu/kermit/
2 April 2002
As of: Kermit 95 1.1.21
This file contains entries from earlier editions of the Kermit 95 Bug List, that were moved to this file when
the bugs were fixed. The numbering is preserved. Bugs fixed since K95 1.1.19
(February 2000) are still in the bug list.
(Fixed in version 1.1.12)
Kermit 95's terminal emulator supports color and highlighting (bold) character
attributes in the Terminal window, which are set by escape sequences from the
other computer. In addition, the underlining attribute is simulated via color
or intensity. This should work well in any size screen, but unfortunately
there is a bug in Windows 95 -- but not in Windows NT -- that prevents the use
of attributes in console screens that are not 80 characters wide and 25, 43,
or 50 lines high (24, 42, or 49 plus Kermit's status line). If an application
attempts to write attributes in such a window, Windows 95 (not just the
application) crashes. Therefore, host-controlled screen-size changing is
disabled by default in Windows 95. You can enable it with the following
command:
SET TERMINAL VIDEO-CHANGE ENABLED
If you do this, then if the host changes the screen size to anything other
than 80 by 25, 43, or 50, then Kermit 95 will refuse to manipulate attributes
in the screen (because if it attempted to change attributes, Windows would
crash). The problem is most commonly visible when switching to 132-column
mode, in which case you probably observe that columns 81-132 are monochrome,
as is new material in the other columns. Switching back and forth between
the Terminal and Command screens in this situation can cause further
confusion. Basically, all bets are off because Windows itself becomes
extremely unstable when an "odd-sized" console window is in use.
If and when the Windows bug is fixed, you can enable screen attributes in
screens of all sizes by issuing the hidden command:
SET TERM ATTR-BUG OFF
Do not do this before the Windows 95 bug is fixed; it will result in a General
Protection Fault and potential loss of data or damage to your system.
Once you have installed the Windows 95 patch or new Windows 95 version (if
any) that corrects this problem, you can put the above command in your K95.INI
file and then you can enjoy full-colored and -highlighted screens of
(hopefully) any dimensions. NOTE: There was no fix for this problem in the
first Microsoft Service Pack, nor in Windows 95 OEM SR2 (Service Release 2).
REMEMBER: This is a Windows 95 bug, not a Kermit bug. You can use any screen
dimensions you like in Windows NT, up to 162 columns and 101 rows.
This is fixed in version 1.1.12.
(NOTE: Users of PCMCIA modems, also see Item 79).
Windows 95 and NT 4.0 only. The interface between Kermit 95 and Microsoft
Telephony does not yet work. It probably can not be made to work until the
conversion of K95 to full-GUI is complete. In the meantime, you can use your
COM devices (COM1, COM2, etc) directly, along with Kermit 95's own built-in
modem support. This has several consequences:
- The configurations you perform on your modem in the Windows Control Panel
will have no effect on Kermit 95. Of course, you should still set up
your modem in the Control panel for other applications that use it.
- Your communication device is set to DEFAULT, rather than a TAPI modem
name. DEFAULT means that Kermit 95 will use the device (COM1, COM1,
COM3, or COM4) that you specified in SETUP.
- The information on the Location and Codes page under the Options menu --
your country code, area code, long-distance dialing prefix, etc -- comes
from answers you gave to these questions in SETUP, since Kermit 95 is not
presently able to obtain this information automatically from Windows.
- If you change locations -- e.g. move to a different city, or take your
laptop on a trip -- you will have to make the appropriate changes in the
Locations and Codes page.
- If you have a type of modem that is not directly supported by Kermit,
you'll either have to use one of the built-in types that is close enough
to work, or else construct a "user defined" type as explained in "Using
C-Kermit", second edition.
- To dial out on a serial device that is owned by TAPI (e.g. that is usually
waiting for a call to come in), you have to shut down the TAPI application
(e.g. your fax reader).
To change your communication device and modem, find the DEFAULT Template (the
entries in the Dialer are listed alphabetically), click on it once with the
mouse to highlight it, then click on the Edit button. This brings up the
Default notebook. Open it to the Communications page and select or add the
desired COM port device name in the "Line:" entry in the upper left. You can
also change the modem selection in the Modem Preferences Type box. Also be
sure to make any other changes or settings you need on this page.
In addition, the notebook for each entry has its own Communications page. So
that you only have to set up your communications device in one place, each of
these pages should have "DEFAULT" (uppercase, without the quotes) in the Line
box, which means that the information about the communications device should
comes from the DEFAULT Template.
However, you can have special settings, and even different communications
devices and modems, for each connection if you want to. In that case, just
choose the desired values on the Communications page for any connection that
is not use the defaults.
Windows 95 (not NT) returns the shifted version of non-alphabetic characters
when Caps Lock is on; for example, if you press the "2" key on a USA keyboard
with Caps Lock on, you get "@" instead of "2". In other words, Caps Lock
behaves like Shift Lock. This is a Window 95 bug. The same K95 code
executing under Windows NT returns the correct values.
Resolution: Fixed in version 1.1.5, in which a new code was added to bypass
the defective Microsoft Windows 95 console agent interface to the keyboard
driver.
In Windows 95, but not NT, buttons and checkboxes on the Dialer notebook pages
might have the wrong background color, resulting in a blotchy appearance, most
notably on the Terminal page. This (purely cosmetic) problem should be
corrected when the next release of our GUI development tool arrives. The
problem occurs with the default Windows color scheme, and apparently does not
occur when certain other color schemes are used.
Also, due to another bug in our GUI development system, it seems that if you
use the Dialer for a long period of time to make lots of changes to the
database, some of the graphic elements begin to deteriorate; for example, the
little arrows on scroll boxes turn into 6's. Workaround: Exit from the Dialer
and restart it.
RESOLUTION: Fixed in version 1.1.5 by a new release of the GUI development
system.
- The Dialer status line's online elapsed-time clock ticker is a bit
idiosyncratic. (Fixed in 1.1.1)
- There might a few cases where the Dialer might not catch Online-to-Offline
transitions or vice versa, and so the Status field might be inaccurate.
(Fixed in 1.1.1)
- The Hangup button appears to have no effect in Windows NT. This should be
fixed shortly by a patch. In the meantime, use Alt-X in the Terminal
screen and HANGUP and EXIT commands in the Command screen to achieve the
same effect in Windows NT.
- Problem
- There is a problem with the down-arrow in the vertical scroll
bar of the main Dialer window, as well as in various other
vertical scroll boxes. If you put the mouse cursor on the
down-arrow button and hold down the mouse button, strange
effects will ensue. Typically, the window will scroll back up
when you let go of the mouse button. Sometimes the down-arrow
mysteriously turns into an up-arrow, and sometimes it becomes
enlarged.
- Diagnosis
- A combination of a Windows 95 bug, in which messages posted to
the event queue can be lost without notification to the poster,
and our GUI development system, whose use increases the number
of messages posted to the event queue.
- Workaround
- Move the mouse off the down-arrow and click it in some other
area. Avoid the problem by using a different method of
scrolling down, such as the down-arrow key, or clicking in the
vertical scroll bar above the down-arrow button.
- Cure
- Fixed in version 1.1.2 by an updated version of our GUI
development system.
- Problem
- ZMODEM downloads fail when K95's FILE COLLISION setting is
BACKUP or RENAME.
- Workaround
- Prior to initiating a ZMODEM download, choose SET FILE
COLLISION OVERWRITE.
- Cure
- This is fixed in version 1.1.1.
- Problem
- ZMODEM transfers do not work on certain TELNET connections.
- Cure
- This is fixed in version 1.1.1.
- Problem
- Alphabetization of Dialer entries is case sensitive.
- Cure
- This is fixed in version 1.1.1.
- Problem
- Sometimes, when launching a connection from the Dialer, K95
says something like "A connection is open on COM1. OK to
Close?" and waits for a reply, even though no connection is
open.
- Diagnosis
- Your K95CUSTOM.INI file, originally generated by the SETUP
procedure, has a SET PORT and SET SPEED command in it for the
port that you chose, if any, in the setup dialog, so that K95
knows the port and speed to use in case you start it directly,
rather than from the Dialer. But if the modem is presenting
the carrier signal, K95 will issue this message.
- Workaround
- Add "set exit warning off" to your K95CUSTOM.INI file, before
the SET PORT command. Unfortunately, this means that you
won't get warnings later on, when you might need them.
- Cure
- Fixed in 1.1.1 by the addition of a new command:
IF STARTED-FROM-DIALER
This is used in the K95CUSTOM.INI file to skip over the
SETUP-generated commands when K95.EXE is started from the
Dialer, and therefore to execute them only when K95 is not
started from the Dialer. NOTE: The patch procedure will not
change your K95CUSTOM.INI file, so if you have this problem,
you'll need to make the change yourself:
- Add:
IF STARTED-FROM-DIALER GOTO MYPREFERENCES
as the first command.
- Add a label:
:MYPREFERENCES
after all the SET MODEM, SET PORT, SET SPEED, and SET DIAL
commands that were generated by SETUP.
- Put all your own customization commands after the
MYPREFERENCES label.
- Symptom
- Specification of the Download Directory in the General page
of a Dialer entry seems to have no effect.
- Diagnosis
- Bug
- Workaround
- Use the "set file download-directory" command in the Command
screen.
- Cure
- Fixed in version 1.1.1.
- Symptom
- Modem customization commands added to K95CUSTOM.INI seem
to have no effect when launching a connection from the Dialer,
even though they do work correctly when when K95.EXE is
started without the Dialer.
- Diagnosis
- The Dialer generates a SET MODEM TYPE command that is
executed by K95.EXE after the K95CUSTOM.INI file is executed.
The SET MODEM TYPE command reinitializes all of the SET MODEM
modem parameters from an internal table, thus overriding the
customizations you made in the INI file.
- Workaround
- Bypass the Dialer if you need to use customized modem
parameters or a user-defined modem type.
- Cure
- Fixed in version 1.1.1.
Also see item 20. Be sure you have modified your K95CUSTOM.INI file as
indicated there, because it can't be done by the patch procedure. Then
add your modem-related commands after the :MYPREFERENCES label. Example
(courtesy Jim Jacobus):
:MYPREFERENCES
; (Fill in your other customization commands here...)
; The following "set modem" commands override the defaults put in by
; the dialer. These are called before dialer. However, the "set modem type"
; prevents these from being overridden as long as the type matches exactly.
set modem type microcom-at-mode
set modem command init-string atQ0X4F0&S0&C1%C3&D2\\K5\{13}
set modem command error-correction on at\\N3\{13}
set modem command hardware-flow at\\G0\{13}
set modem command software-flow at\\G1\{13}
- Symptom
- When a TAKE command, or SET DIAL DIRECTORY, or SET NETWORK
directory command is given to K95.EXE, specifying a file
without including the full path, sometimes K95.EXE can't find
the file.
- Diagnosis
- K95.EXE 1.1 tries to open the file with exactly the name that
was given. If that fails, and if the filename did not include
a full path, it looks in an alternate location, depending on
the command: the TAKE command looks in the SCRIPTS
subdirectory of the Kermit 95 directory; the SET DIAL
DIRECTORY and SET NETWORK DIRECTORY commands look in the
PHONES subdirectory. However, since key mapping files are
kept in the KEYMAPS subdirectory, the TAKE command can't find
them if a full pathname is not given.
- Workaround
- Specify a full pathname.
- Cure
- The TAKE command has been modifed in version 1.1.1 to use the
following method to find its file:
- Use the name as given.
- If that fails, look in the Kermit 95 directory.
- If that fails, look in the SCRIPTS subdirectory.
- If that fails, look in the KEYMAPS subdirectory.
- Symptom
- When a Dial Prefix is set in the Dialer's Options ->
Dialing -> Location and Codes page, e.g. to disable call
waiting, it is mistakenly taken to be both the prefix and
suffix. Also, any Dial Suffix specified in the Dialer is
ignored.
- Diagnosis
- Bug.
- Workaround
- Put a SET DIAL PREFIX command in your K95CUSTOM.INI and leave
the Dial Prefix blank in the Dialer.
- Cure
- Fixed in 1.1.1.
- Symptom
- Certain numbers, like maximum redials, were restricted by the
Dialer to insufficient ranges.
- Diagnosis
- Bug
- Workaround
- Enter the big numbers anyway.
- Cure
- Fixed in 1.1.1.
- Symptom
- Backspace key, used when connected to the Kermit BBS,
does not work as expected.
- Diagnosis
- Bug
- Workaround
- Change the Backspace key setting in the Kermit BBS notebook
entry to make it send Backspace rather than Delete.
- Cure
- Fixed in 1.1.1.
- Problem
- ZMODEM downloads in recovery mode fail if Kermit 95's FILE
COLLISION is set to UPDATE.
- Diagnosis
- An fundamental difference between Kermit and ZMODEM protocols.
- Workaround
- Change Kermit 95's FILE COLLISION setting to BACKUP or
OVERWRITE, or use Kermit protocol for recovery instead of
ZMODEM protocol.
- Cure
- Fixed in 1.1.5.
Fixed in 1.1.2.
Prior to 1.1.2, Kermit's DIAL command would disable autoanswer. But there is
no reason why this must be done, so beginning with version 1.1.2, the DIAL
command never disables autoanswer.
The Compose key (Alt-C, verb \Kcompose) was inoperable in 1.1 and 1.1.1 due to
a timing problem. This was fixed in 1.1.2. OE/oe was also added to the
compose-key table to send the DEC MCS codes for upper- and lowercase OE
ligature. (Note, however that the OE ligature still can't be displayed
correctly in most cases, since it is lacking from code pages 437 and 850.)
- Problem
- When Kermit 95's TERMINAL CHARACTER-SET was TRANSPARENT and
its emulation was VT-100 or higher (but not ANSI), it
erroneously treated characters 0x80 - 0x9F as C1 control
characters instead of graphic characters.
- Cure
- Fixed in 1.1.2.
Fixed in version 1.1.2:
- VT100 alignment pattern (DECALN) now follows default coloration.
- CSI ? 5 h / l (DECSCNM) now preserves boldness across screen reversals.
- VT52 submode identify query response fixed.
- Improper response to CSI6x;y"p, resulting in spurious switch to 7-bit mode
- and/or VT220 emulation.
NOTE: With these fixes, Kermit 95 obtains a perfect VTTEST score with two
exceptions: (1) the host can't control the keyboard autorepeat feature; (2)
in Windows 95 only (not Windows NT), due to a bug in the Windows 95 console
driver, video attributes of 132-column screens are incorrect.
Fixed in version 1.1.2:
- Alt-O (\Kprtauto) changed to autoprint, rather than transparent print.
- Transparent print was leaking characters on the right screen margin.
- No more sending Ctrl-Z at end of print job.
Fixed in 1.1.2.
XMODEM uploads would fail if the packet length were not explicitly set to
128 after choosing XMODEM protocol. Fixed in 1.1.2.
XMODEM downloads would crash Kermit 95 under certain conditions. Fixed in
1.1.2.
MOVE and MMOVE are like SEND and MSEND, except they are supposed to delete
the (each) source file after it is transferred successfully. When Kermit 95's
protocol was set to XMODEM, YMODEM, or ZMODEM, Kermit 95 would fail to delete
the source files. Fixed in 1.1.2.
Fixed in 1.1.2:
- Ctrl-C can now interrupt file transfer at any point.
- Ctrl-C can now interrupt server mode.
A file transfer that was interrupted with Ctrl-C could not be recovered (with
Kermit or ZMODEM protocol) in the same session because its handle was not
released, thus preventing the same file from being accessed a second time.
Fixed in 1.1.2.
Fixed in 1.1.2 -- now the download directory is used when Kermit 95 receives
files whie in server mode, and also when it is the client and is given a GET
command. Previously the download directory was used only with the RECEIVE
command.
In any command where a filename was to be parsed, such as SEND, TAKE, etc,
a variable (such as \%f) could not be used in place of the filename. Fixed
in 1.1.2.
On the Dialer's terminal page for any given connection, if you chose a
terminal character set whose name contained a dash, it would fail to take
effect. Fixed in 1.1.2.
Fixed in 1.1.2.
Fixed in 1.1.2.
In Windows 95, but not Windows NT, the serial communications driver never
sends an XON after once having sent an XOFF. This will cause flow-control
deadlocks. Workarounds: (a) Enter the terminal screen and type Ctrl-Q to
clear the deadlock; (b) Don't use Xon/Xoff flow control. This problem will
disappear when Microsoft fixes the driver. Fixed in 1.1.11.
Fixed in 1.1.2. And in 1.1.3 the name is LOGIN.KSC.
When Kermit 95's protocol is set to XMODEM, YMODEM, or ZMODEM, the "as-name"
given in a SEND or RECEIVE is ignored (except in XMODEM receives, where it
is not only not ignored, it is required). Currently, this is a limitation of
the software. (Fixed in 1.1.5)
Fixed in 1.1.2.
Bug. Fixed in 1.1.3. Now it should work for XYZMODEM transfers too.
When pasting a large amount of text, e.g. more than 100 lines, into the
terminal screen on a serial connection, Kermit 95 could crash. Bug.
Fixed in 1.1.3.
- Problem
- Giving the SET FILE DOWNLOAD-DIRECTORY command without specifying
a directory name is supposed to make incoming files go to the current
directory (in the absence of any other instructions), but it didn't.
- Diagnosis
- Bug.
- Workaround
- Use SET FILE DOWNLOAD-DIRECTORY . ("." is means "current directory").
- Cure
- Fixed in 1.1.3.
Bug. Fixed in 1.1.3.
Bug. Fixed in 1.1.3.
Bug. Fixed in 1.1.3.
- Symptom
- Temporary files are always created in the TMP subdirectory of
the Kermit 95 directory. This causes trouble when Kermit 95
is being run from a central server (at legitimately bulk-licensed
sites).
- Cure
- Beginning with version 1.1.3, Kermit 95 can be configured to create
temporary files in any desired location, as specified by the user's
K95TMP, TEMP, or TMP environment variables (if any, in that order).
- Diagnosis
- Typo in EMACS.INI file. The SET KEY command for Alt-y used the
Alt-x's scan code.
- Workaround
- Change "set key \2168 \27y" to "set key \2169 \27y" in EMACS.INI.
- Cure
- Fixed in 1.1.3.
The user should be able to cancel an INPUT command by "pressing any key", but
this did not work. INPUTs could only be canceled with Ctrl-C, which would
also cancel the entire script. Fixed in 1.1.3 -- now pressing any key (such
as space bar) will cause the INPUT to fail immediately, but will not terminate
execution of the script. Also see next item.
Bug. Fixed in 1.1.3. These commands were added for compatibility with MS-DOS
Kermit, where they are primarily used as a way to tell whether an INPUT
command failed because it timed out or because the user interrupted it from
the keyboard. K95 1.1.3 now has a better way to determine why an INPUT
failed: the new \v(instatus) variable. See DOCS\UPDATES.TXT or
"Using C-Kermit", second edition, for details.
Fixed in 1.1.3; \v(cpu) now returns your actual CPU type, such as
"intel-pentium" or "ppc-604".
Bug. There is no such escape sequence for real VT terminals. The code that
handled this escape sequence has been removed in 1.1.3.
Transparent printing is supposed to route incoming characters straight to
the printer without any filtering or modification. Kermit 95, however, was
performing character-set translation on incoming text before sending it to
the printer. Fixed in 1.1.3. Character-set translation is still done for
autoprinting (copy screen lines to printer), as it should be.
Bug. Improper modem commands were issued for flow control and certain other
parameters on the "microcom-at" modem type. Fixed in 1.1.3.
Bug. Fixed in 1.1.3.
Bug. Fixed in 1.1.3.
Bug. Fixed in 1.1.3.
This type of warning is issued when you try to exit from Kermit 95 or change
communication devices, and Kermit 95 thinks you still have a connection open.
If your current communication device is COM1 (or COM2, etc), and the carrier
signal is NOT detected on that device, then the message should not be issued.
An improper test for carrier was being done. Fixed in 1.1.3.
SET PORT, when given without a port name, closes the current connection but
does not open a new one. This is necessary sometimes, as when you are using
Kermit 95 to execute a script locally, without communicating, and you don't
want to needlessly occupy a port, which would prevent other processes from
getting access to it. This works, but Kermit 95 erroneously says "Sorry,
access to device denied: 0" and sets FAILURE for the command. The error
message (and FAILURE condition) were removed in edit 1.1.3. Note that when
Kermit 95 has no communication device open, SHOW COMMUNICATIONS reports the
port as "0".
This is a subtle timing problem, reported by only one user. A technical
explanation is beyond the scope of this document, but an effective workaround
was found: change the timing.
When Kermit 95 dials, it takes your DIAL TIMEOUT value and subtracts a certain
number from it, and gives the result to the modem, e.g. via ATS7=xxx. This
way the modem should time out before Kermit does, and return (e.g.) NO CARRIER
connection is not made. This allows Kermit to handle the situation "normally"
and in some cases to actually determine the failure reason from the modem.
However, should the modem fail to respond within Kermit's timeout interval
(which is normally 10 seconds longer), then Kermit has to break out of its
dialing routines, kill some threads, reset some signal handlers, etc.
It turns out that under certain very unusual conditions, these two actions can
collide or conflict in some way that can cause Kermit to crash. Until a solid
cure is found, the following workaround is available as of version 1.1.3:
SET DIAL TIMEOUT seconds [ differential ]
If you ever experience an access violation as a result of a DIAL command
timing out, specify a different differential. Normally it's 10. Try 5
or 15 or some other number. You can even use a negative number here to force
Kermit to time out BEFORE the modem times out, which seems to be a solid cure
for this problem, except that the desired status reports from the modem are
never seen.
In the Kermit 95 SETUP program, in the section where you choose a modem
type, the "Practical Peripherals" choice was erroneously listed twice. The
first instance (at the bottom of the first column) really should have been
"Multitech". Fixed in 1.1.3.
On a real VT terminal, Ctrl-Space sends NUL (ASCII 0). In Kermit 95, it
was sending Space. Fixed in 1.1.3. Now Ctrl-Space, Ctrl-2, and Ctrl-@ all
have a default definition of \Knull (ASCII 0).
Restriction. Although TAKE, SET DIAL DIRECTORY, and certain other commands
might search through an internally defined list of directories for their
files, this is not presently done for filenames given on the command line.
Fixed in version 1.1.8.
Bug. Fixed in 1.1.3.
On systems with national keyboards and drivers, some features might not work
as expected in Kermit 95 under Windows 95, even though they work perfectly
in Kermit 95 under Windows NT. For example, dead-key combinations such as
"^" followed by "A" to make A-circumflex with the German keyboard and driver.
This seems at first to be a Kermit problem, because it does not show up on
the same system at the DOS prompt or using DOS applications.
- Diagnosis
- Unlike COMMAND.COM and the other DOS applications, Kermit 95 reads
scan codes, not "bare characters", from the console (keyboard driver), and
Windows 95 console support is full of bugs. Unfortunately, this is a problem
that only Microsoft can fix. It did not elect to fix this or any other
console-related bug in its first service pack.
- Workaround
- Use Kermit's Compose key. For example, use Alt-c, then ^, then A
to enter A-circumflex.
- Cure
- We will escape from all of these console-mode bugs when we convert K95
to a full GUI program. We are working on that now.
- Resolution
- Version 1.1.5 bypasses the buggy Windows 95 drivers.
Symptom: Misformatted screens in VT220/320 applications that use the ECH
sequence: CSI n X, meaning: erase n characters starting at current position,
effectively turning them into spaces. Bug in Kermit 95 1.1.2 and earlier.
Fixed in 1.1.3.
- Restriction
- the file transfer display code and statistics displays of the
XYZMODEM protocol code are not yet hooked in with those of the Kermit protocol
code. This will be remedied in a future release.
- Resolution
- Remedy is new code in 1.1.5.
SYMPTOM: The selected terminal colors (if they are not the default blue
on light gray) are not effective on the initial Terminal screen.
Instead, the initial terminal screen has the default background color
rather than the selected one. All new material has the right coloration,
and as soon as your host service or application sends a "clear screen"
escape sequence, or the first screen's worth of material scrolls away,
the problem resolves itself, so only the first screen looks funny.
Diagnosis: Bug. Fixed in 1.1.4.
Symptom: Certain host applications were coded to take advantage of MS-DOS
Kermit's detailed strategies for handling colors. These applications might
not work as expected with Kermit 95. Cure: a new command, SET MSKERMIT
COLOR, was added in 1.1.4 to select between the two styles of color handling.
(NOTE: THIS COMMAND REMOVED in 1.1.5 and replaced with several new commands.)
See DOCS\UPDATES.TXT for details.
When Kermit 95 sends transparent-print or autoprint material to the printer,
the Windows printer driver is bypassed entirely, and plain text is sent to the
printer. This works OK with printers that can print plain text, but it does
not work, for example, with PostScript printers. Version 1.1.4 includes a new
command to let printer output be redirected through any desired input/ output
filter, and such a filter, TEXTPS.EXE, is supplied for converting plain text
to PostScript.
When dialing out through a modem on a "reverse terminal server", neither
REDIAL nor DIALing again would work unless another SET HOST command was given.
Fixed in 1.1.4.
The command-file line displayed by SET TAKE ECHO ON were not always reset when
starting a new command file. Fixed in 1.1.4.
IF FAIL or IF SUCCESS tests were ineffective after DEFINE or ASSIGN commands,
which always succeeded, even if the definition was too long for Kermit's
internal buffer. In 1.1.4 these commands properly fail if the definition is
too long.
The temporary files created by the Dialer to launch to launch K95.EXE on a
certain connection were given the extension .SCR (script), which conflicts
with Windows 95's use of .SCR to denote a screensaver. Beginning in version
1.1.4 they are created with extension .KSC (Kermit script).
Thus, if you save one of these files (using SET STARTUP-FILE KEEP in K95.INI),
and if you create an association between .KSC files and K95.EXE, you can
launch K95.EXE on a particular connection simply by clicking on the KSC file.
You can create the association in Explorer -> View -> Options -> FileTypes.
Call it "Kermit script" and Associate Open with 'c:\k95\k95.exe -C "take %1"'.
The MSEND command, which sends a list of files in a single operation,
failed to process the second and subsequent file specifications in the list
if the protocol was not Kermit. This is fixed in version 1.1.4 so that now
MSEND can be used with YMODEM or ZMODEM protocol. It can't be used with
XMODEM because XMODEM, by definition, can transfer only one file at a time.
Bug. Fixed in 1.1.4.
VT220/320 string-bearing escape sequences, of the sort that end with String
Terminator (ST), were parsed erroneously if the data field happened to contain
control characters, e.g. inserted by VMS terminal line wrapping. The control
characters should be sent to the screen and otherwise ignored. If a control
character was inserted between the ESC and \ of a 7-bit ST, the emulator would
hang because it did not recognize the ST. This affected transparent print
also, even though the emulator is supposed to ignore all escape sequences up
to CSI 4 i (but it still has to parse them). Fixed in 1.1.5.
Even after the color-related fixes in 1.1.4, there were still problems when
reverse-video cells (or screens) had material erased from them by host
escape sequences; the cells would be cleared using the foreground rather than
the background attribute. Fixed in 1.1.5.
Bug. Fixed in 1.1.5.
When the VT emulator was in DECNRCM (7-bit ISO 646 National Replacement
Character-set Mode), and the NRC set was designated via any of the standard
DEC / ISO-2022 escape sequences, the set was assigned only to the single area,
G0..G3, indicated by the escape sequence, and not to all four of G0..G3, as
required by DEC terminal specs.
Autodownload did not work if the terminal character-set was any of the
7-bit ISO 646 national versions (Canadian-French, etc). Fixed in 1.1.5.
Fixed in 1.1.5 by a new release of our GUI application development system.
The Dialer only allows one active session per entry. This is simply so that
it will know which process should receive the Hangup signal when you click
the Dialer's Hangup button. Workaround: If you want to start multiple
sessions to the same host from the Dialer, use the Clone button to create
multiple entries that are identical, but with different names, like "myhost1",
"myhost2", "myhost3", etc. Fixed in version 1.1.8.
Keyboard shortcuts for Dialer buttons and notebood tabs were added in
version 1.1.5.
The Dialer would crash if you edited an entry name and then selected the
entry. Fixed in 1.1.5.
All of the PBX strings were mushed together into a single string.
Fixed in 1.1.5.
Fixed in 1.1.5.
Microsoft gethostbyname() calls DNS to resolve the hostname even if it is
already a numeric IP address. If DNS is unavailable or not working right
(which is probably why you used a numeric address in the first place), this
causes unacceptable delays, or fails outright. Fixed in 1.1.5 by not calling
gethostbyname() if the address was already numeric.
Various minor errors in Practical Peripherals, Multitech, and AT&T Paradyne
modem support (mostly due to incompatibilities in new models) fixed in 1.1.5.
If you used filename completion in a SEND command, the SEND command tended
to fail. Diagnosis: Bug. Fixed in 1.1.5.
Fixed in 1.1.5. Backslashes can now be used in the normal in OUTPUT
commands, e.g. "output \\" sends one backslash.
There was a bug preventing transmission of more strings longer than 80
characters, fixed in 1.1.5.
If a session log was active during an INPUT command, all incoming characters
were recorded twice in the log, and character sequences could sometimes appear
out of ourder. Fixed in 1.1.5.
The fixed 256-byte size of the INPUT buffer was too small for some scripting
applications. In version 1.1.5, the size can be selected with the new SET
INPUT BUFFER-LENGTH command.
Kermit 95 1.1.4 failed to correctly evaluate a filename if it was composed of
a mixture of literal characters and variable names. Fixed in 1.1.5.
Fixed in 1.1.5. See DOCS\TEXTPS.TXT.
Bug. Fixed in 1.1.5.
Bug introduced in version 1.1.5. Workaround: set your terminal type to VT102,
VT220, or VT320. Fixed in 1.1.6.
The multilevel keyboard maps that were introduced in version 1.1.5 resulted in
SET KEY commands taking longer than before. Thus files that contained many
SET KEY commands could take a long time to execute, sometimes 30 to 60
seconds. SET KEY has been accelerated in version 1.1.6.
Bug, fixed in version 1.1.6.
Constructions such as "open read \&@[5]" would fail due to a bug introduced
in the fix to bug 129, in version 1.1.5. The new bug is fixed in 1.1.6.
The layout of keyboard modes like EMACS, Hebrew, and Russian can not be
modified at all, period. Also, the layout of the terminal-dependent keymaps,
such as VT220, Wyse, etc, can not be modified without disabling automatic mode
switching for the affected keys. This is a restriction of the new multilevel
key mapping feature of version 1.1.5. You can always get around it with SET
KEY (or files or macros full of SET KEY commands). A more flexible approach
is under consideration for future releases.
This is fixed by the addition of the SET TERMINAL KEY command in version 1.1.8.
The SCOANSI, AT386, and Avatar emulations, which were added in version 1.1.5,
as well as the original ANSI (ANSI-BBS) emulation, always issued a complaint
if your code page was not CP437. In version 1.1.6, the complaint is not
issued. Whatever your code page is -- 850, 852, 862, 866 -- it is used as the
local end of your terminal character-set. Every code page includes a full
complement of single-line and double-line box characters, as well as solid
and shaded boxes of various shapes. The double-line-to-single-line characters
that are found in CP437 are not included in CP850 or CP852, but hopefully most
applications avoid these characters.
The SCOANSI and AT386 terminal types did not properly switch among G1-G3
in response to CSI 10/11/12 m escape sequences. Bug, fixed in 1.1.6.
The GUI registration program that runs during SETUP stores your name and
company name in Latin-1 (or Latin-2, etc, depending on your locale). If it
contains accented or non-Roman letters, they appear correctly in GUI programs
(such as the Dialer) but they appear incorrectly in K95.EXE, because the
console window uses a PC code page and not "Windows Latin" character sets.
This is problem is fixed in K95.EXE 1.1.6 by translating the characters to
the current code page.
You can't use ESC or TAB in the command parser to complete a directory name,
as you can to complete a filename or a keyword. Fixed in 1.1.12.
DIRECTORY xxx, where "xxx" is the name of a nonexistent file or directory,
gave a spurious "invalid command" message, instead of a "no files match"
message. Also, "directory xyz" did not work in K95 as it does in DOS.
Both fixed in 1.1.6.
Bug, fixed in version 1.1.6. Also, incorrect placement of some messages on
the file transfer display, also fixed in 1.1.6.
Bug, fixed in version 1.1.6. Also, the command did not wait for confirmation;
also fixed in 1.1.6.
Bug, fixed in 1.1.6.
Bug, fixed in version 1.1.6.
Bug, fixed in version 1.1.6.
File Transfer Settings page grayed out fields that are now legal choices
for XYZMODEM protocol. Fixed in version 1.1.6.
Bug, fixed in version 1.1.6.
XYZMODEM file transfer commands (SEND, RECEIVE, etc) always returned the
same status code, regardless of whether the transfer succeeded or failed.
Fixed in 1.1.6. Now IF FAIL / IF SUCCESS works with these protocols too.
Beginning in version 1.1.6, they do.
On new connections that are opened after the first connection is closed.
Bug. Fixed in 1.1.6.
Data could be lost during uploads on high-speed serial connections. This is
partially due to a bug that first appeared in version 1.1.5, fixed in 1.1.6,
and partially because the default flow control for serial connections has
always been whatever the port device driver had been set to. In 1.1.6, the
default flow control has been changed to AUTOMATIC, which picks the most
appropriate and effective type of flow control for each connection,
automatically.
Prior to version 1.1.6, Kermit 95 could not send (or otherwise read) a
file that was open, either by another process or even by itself. Now files
that are to be accessed read-only are opened in shared mode.
The algorithm used for calculating timeouts during Kermit file transfer,
introduced in version 1.1.5, was too conservative. In version 1.1.6, the
values should be more realistic. Also, now you can specify the minimum
and maximum timeout values: SET SEND TIMEOUT DYNAMIC .
Bug. Fixed in 1.1.6.
No matter how it was invoked, it would only print its help message.
Bug. Fixed in 1.1.6.
The percent efficiency that is shown after a file transfer is, in general,
meaningless on dialup connections, since Kermit only knows the interface
speed, but not the actual connection speed. Therefore in 1.1.6, the
efficiency is not shown except on direct serial connections, where the
connection speed is known.
Bug, fixed in 1.1.7.
Because of a Windows 95 bug (Entry 2 in this file), "odd" screen sizes are
written without attributes, such as colors. SCOANSI and AT386 emulation do
not have a status line, therefore use 25 lines. But the Dialer would show 24
unless you changed it by hand, and 24 is an "odd" size, and so SCOANSI and
AT386 screens would come out monochrome, and fractured in several other ways.
Fixed in 1.1.7. The Dialer now adjusts the other paramaters on the Terminal
page appropriately for any emulation you select. Of course you can then
readjust the parameters separately after that. Also, Kermit 95 now takes into
account whether there is a status line when deciding whether the screen is
an "odd" size.
Bug. Fixed in 1.1.7.
Bug. Fixed in 1.1.7.
A change in the latest release of our GUI development system. Worked around
in 1.1.7.
Bug. Fixed in 1.1.7.
It should have been Latin-1. Transparent is only for ANSI terminal types.
Using Transparent with VT emulation results in badly fractured screens when
using VMS applications that send C1 control characters. Fixed in 1.1.7 by
having the Dialer automatically set Latin-1 when VT220, VT320, or WY370
terminal type is chosen.
Under some circumstances, bad screen coordinates or size information can be
saved, and then they are used when the Dialer is next started. Workaround:
Maximize the Dialer, change the View options, exit from the Dialer, start the
Dialer again. Version 1.1.8 makes additional checks upon startup.
The IP address changed. The new one is ics.onenet.net:5000. Fixed in
version 1.1.8.
The original file was always overwritten. Fixed in 1.1.8.
Bug, fixed in 1.1.8.
When you press Scroll Lock, the screen and keyboard stay locked forever;
pressing another key does not release it. Bug. Workaround: SET KEY \401
\Kignore. Fixed in 1.1.8.
Fixed by addition of SET TERMINAL KEY command in version 1.1.8.
CTRL2CAP.TXT says to add the "device=" statement to the [enhanced] or [386enh]
section of WINDOWS\SYSTEM.INI. In fact, it should go in the [386enh] section.
LOG SESSION translates character sets and C1 control characters before writing
them to the log. This prevents proper replay of the log. Bug. Fixed in
1.1.8.
VT terminal ICH function is implemented for VT220 and 320, but not 102.
Bug. Fixed in 1.1.8.
The VT emulator treats CSI ? ; 2 h as CSI 0 ; 2 h. This can result in a
keyboard lock condition in at least one emulation. Bug. Workaround: hit
Alt-R (\Kreset) to reset the emulator. Fixed in 1.1.8.
VT emulator does not preperly handle handle sequences for DEC Alternate ROM
1 (ASCII) and Alternate ROM 2 (Dec Technical), resulting in unpredictable
character-set selection. Fixed in 1.1.8.
VT emulator Secondary and Tertiary DA Reports (DA2 and DA3) are treated as
Primary DA. Bug. Fixed in 1.1.8.
DECID returns the wrong seqence when executed in a VT mode greater
than VT102. Bug. Fixed in 1.1.8.
Fixed in 1.1.8.
On NT systems, if the screen buffer had dimensions H x W > 8192, as when the
user sets up a large scrolling virtual console window, Kermit 95 crashes when
started. This is because K95 only supports an active screen of up to 8192
cells. Workaround: Don't create huge windows; use K95's built-in scrollback
feature instead (even though it doesn't have scroll bars). NT users should
not start K95 in a window larger than H x W = 8192, or with more than 100
lines, or more than 155 columns. Version 1.1.8 checks these conditions and
handles them gracefully.
There can be only one active LAT session at a time, even if you are running
multiple copies of K95. Bug. Fixed in 1.1.8.
Fixed in 1.1.8.
Fixed in 1.1.8.
There is no way to specify an unlimited or infinite timeout for INPUT or
MINPUT. Fixed in 1.1.8: 0 = no timeout, -1 = infinite timeout.
INPUT and MINPUT are documented as running all sequences through the
emulator. However, APC and autodownload sequences are ignored.
Fixed in 1.1.8.
Such as "echo My current directory is C:\K95\". Backslash, when used at
the end of a command, is a little-known synonym for "-", the command
continuation character. In 1.1.8, backslash was eliminated as a continuation
character; only "-" can be used for continuation.
Removed in 1.1.8.
The Close box [x] does not make the K95.EXE window go away without putting up
a warning box first (Windows 95 only). This is a Yet Another Feature of
Windows 95 in its treatment of Console windows. If you use the close box to
shut down K95, and answer OK in the warning box, Windows terminates K95
immediately without passing a Close Application message to it, and thus K95 is
completely unable to release any resources that it might have in its
possession such as a TAPI modem handle. K95 can do nothing to either disable
the close box or trap the event, so all we can tell you is: don't use it.
Exit from K95 by giving it an EXIT command at its prompt, or by using the
\Kquit verb (assigned to Alt-Q by default), or by using the SET EXIT
ON-DISCONNECT option if you prefer that.
Worked around in 1.1.19, in which a way was found to disable the Close box.
If the statusline was disabled with SET TERM STATUSLINE OFF or via use of a
terminal type that requires it be off (SCOANSI,AT386), then in a Telnet session
to a host that supports NAWS, the size of the screen would be reduced by one
line each time the terminal size was reported to the host. Fixed in 1.1.8.
The SET TERM CHARACTER-SET TRANSPARENT command results in page faults later on
during program operation. Workaround: Use SET TERM REMOTE-CHARACTER-SET and
SET TERM LOCAL-CHARACTER-SET instead. Fixed in 1.1.9.
Workaround: SET TERMINAL UNICODE OFF. Fixed in 1.1.9.
Transparent printing, print-screen, and print-line operations were printing
the wrong data. Symptom: only the left half of the screen would be printed,
perhaps with garbage characters inserted between each real character, and the
right half of the screen would not be printed at all. An artifact of the
Unicode conversion. Fixed in 1.1.9.
Dead-key and Alt-Gr broken combinations stopped working again in version
1.1.8. Another artifact of the Unicode conversion. Fixed in 1.1.9.
APC command handling and character set translation during the INPUT command
were not working. Fixed in 1.1.9.
K95's terminal type is set (redundantly) the first time it is sent to the
host in a Telnet terminal-type negotiation. No harm done. Fixed in 1.1.9.
CD \xXXXXXX, where \xXXXXXX is the name of a directory that does not exist,
and x is one of the backslash-code introduction letters (such as k), can
produce unpredictable results. Workaround: Use forward slash (/) instead of
backslash as the directory separator. Or use double backslashes. (Also see
Bug 174.)
Host-based WordPerfect (e.g. on AIX), when used in VT320 mode, was sending a
"made up" escape sequence, ESC ) ?, to designate CP437 to G1 -- something
obviously not supported by a real VT320, nor by ISO 2022 rules. However,
since this worked with MS-DOS Kermit, we now allow in K95 too, as of K95
version 1.1.9.
Bug. Fixed in 1.1.9.
Fixed in 1.1.9.
Fixed in 1.1.9.
Fixed in 1.1.9 -- Now, if the Dialer detects a corrupt entry, it gives you
a choice of deleting the bad entry or leaving it in the database. If the
entry is kept, it is ignored.
Bug, fixed in 1.1.9.
Bug, fixed in 1.1.9.
LOGIN.KSC, Kermit 95's "universal login script", was modified for version
1.1.8 to allow for the possibility that the user ID prompt might not appear if
TELNET NEW-ENVIRONMENT or RLOGIN protocol was used, but the new version failed
to work when launched from the Dialer and a password was included. Fixed in
version 1.1.9, LOGIN.KSC V3.0.
If you build up a list of files to be sent using ADD SEND-LIST and then
give a SEND command with no operands, it sends the files in the send list.
If you give a MOVE command with no operands, it does the same thing. But
unlike SEND, MOVE should delete each original after sending it successfully;
fixed in version 1.1.9.
Bug, fixed in version 1.1.9.
If TERMINAL SEND-DATA is OFF and the host requests the terminal to send
data, Kermit 95 1.1.8 would simply not send it. But this tended to confuse
the host application. Kermit 1.1.9 sends the data, but replaces the actual
data characters with blanks.
When sending a file with a window size greater than 2, Kermit 95 could send
data packets beyond the end of its window if acknowledgements were received
out of order when the window was blocked AND there was a long round-trip
delay. Bug, fixed in version 1.1.9.
The VMS command SET TERMINAL/WIDTH={80,132} sends a screen-size changing
escape sequence, which is supposed to change the screen size AND clear the
screen. Kermit 95 was not clearing its screen when the new size was equal to
the current size. Bug, fixed in 1.1.9.
When the first command-line argument to K95 is a filename, K95 should execute
the commands from the file, e.g.
k95 filename
But if the filename contains spaces, this doesn't work in version 1.1.8,
even if the filename is properly quoted to the shell:
k95 "this is a file name"
Workaround: Don't use files that have spaces in their names for this purpose.
In fact, filenames with spaces are a bad idea to begin with and confuse many
programs, not just K95. This bug is fixed in version 1.1.9.
In version 1.1.8, if K95 was started with a command file as its first
command-line argument, and the command file was not given a full pathname,
then under certain circumstances, the command screen would be corrupted or
frozen. Most typically this would happen when starting K95 from a shortcut.
Bug, fixed in 1.1.9.
If your logical screen is wider than your physical screen, and the cursor
is placed one column to the right of the rightmost visible column (where
you should not be able to see it), a "ghost" cursor will appear at some
other position on your screen. This is a cosmetic bug only. If you
position the window to show the part of the screen where the cursor really
is, it will appear in the correct place. Fixed in 1.1.10
During the development of 1.1.9, it was discovered that the dynamic changing
of console window sizes in Windows 95 (not NT) was much more dangerous than we
ever imagined, due to the Windows 95 bug noted in the BUGS.TXT file, item 2.
If the width is not 80 and the height is not 25, 43, or 50 (24, 42, or 49 plus
status line), screens could become jumbled, confused, or frozen; Kermit could
crash; Windows 95 itself could crash. This could happen even with our "ugly
workaround" to the Windows 95 bug, namely not writing any attributes into
"odd-sized" screens (because if we did, there would be an invalid page fault
in the Windows 95 kernel).
So to prevent the corruption of Kermit 95 or Windows 95, the default for the
SET TERMINAL VIDEO-CHANGE command was changed from ENABLED to DISABLED. Thus
sessions that you had previously set up to have long and/or wide screens now
have 24x80 screens, and a new distinction was made between the "logical
screen" (e.g. 24x132) and the "physical screen" (e.g. 24x80). Thus you could
scroll the physical screen horizontally back and forth across the logical
screen, and there were even new visual cues in the status line to this effect.
So far so good.
UNFORTUNATELY, this concept was also applied -- mistakenly -- to the vertical
screen dimensions as well as the horizontal ones. So if you started a
sessions with (say) a 42-line screen (43 including the status line), and with
VIDEO-CHANGE disabled (the new default), the bottom 18 lines of the logical
screen would be hidden from view, and the cursor would be stuck in the status
line, resulting in an essentially useless terminal session.
RESOLUTION: In version 1.1.10, a new command was added:
SET TERMINAL VIDEO-CHANGE WIN95-SAFE
and WIN95-SAFE was made the default for SET TERM VIDEO-CHANGE. When this
setting is in effect, your TERMINAL HEIGHT or COMMAND HEIGHT may be set to 25,
43, or 50 lines (24, 42, or 49 if your terminal emulation window has a status
line). You may set your TERMINAL WIDTH to any reasonable value, but this will
become its logical width, not its physical one, and you can use horizontal
scrolling to view portions that are not visible.
In the Dialer, if you choose a terminal screen height that is not 24, 42, or
49 (plus status line), or 25, 43, or 50 (without status line), then a new
alert box will warn you about the situation. Of course, you have to go out
of your way to select an unsafe height to begin with...
If you make a TES32 connection to a VMS system but then you don't log in
within VMS's timeout interval, K95 can crash when VMS disconnects the TES
session. Bug. Fixed in Kermit 95 1.1.10.
There were some bugs in HP terminal emulation, fixed in version 1.1.10.
Mostly involving the application of character attributes and line wrap.
Now the HP emulator works satisfactorily with SAM (the HP system
administration tool) in terminal mode.
Session logs were recording data being sent to the host when even though local
echo was off. Bug, fixed in 1.1.10.
In version 1.1.9 and earlier, the VT100 and VT102 numeric keypads were not
mapped to the PC numeric keypad by default (even though the VT220 and VT320
keypads were). Fixed in 1.1.10.
In 1.1.9, the size of the console window was not being restored to its initial
conditions under some circumstances, and if the height of the terminal screen
was greater than that of the command screen and the width was greater than
that of the command screen, then the command screen would be set to the wrong
size. Bugs, fixed in 1.1.10.
If the host sent an escape sequence to turn off the cursor, the SET TERMINAL
CURSOR did not turn it back on. Bug, fixed in 1.1.10.
The full path was being sent when only the file name should have been used.
Bug, fixed in 1.1.10. But note the following "gotcha" in XMODEM and YMODEM
protocols: If the file name contains an uppercase letter 'C' or 'G', then the
wrong CRC check algorithm will be used if the name is echoed back from the
receiver. There is nothing Kermit 95 can do about this.
The following bugs were fixed in 1.1.10: DG210 was sending the same terminal
ID as DG200; DG emulations were sending ID in wrong format; TELNET terminal
type for DG200 and DG210 should have been D200 and D210; the default
assignment for the Enter key was CR when it should have been LF.
The commands to set the default editor (for the K95 EDIT command) used
hardwired directory and filenames that might be wrong for some PCs, especially
in Windows NT. Fixed in 1.1.10 by using more flexible definitions suggested
by Patrick Russell, K95.INI V2.1.
The HOSTMDM.KSC script, which waits for incoming host-mode calls, would
override any SET MODEM commands you might have given in your K95CUSTOM.INI
file; this was particularly serious for user-defined modems, since it would
wipe out the entire definition. Fixed in 1.1.10, HOSTMDM.KSC V2.0.
On some Windows 95 systems, but not others, the new K95CINIT program pops up
an error box containing the message "Port COMx, Status: Unable to
SetCommStat", which must be closed manually. After the box is closed, the
port is usable, so the utility is working. The problem is caused by bugs in
the Windows APIs. Fixed in 1.1.11 by issuing a different K95CINIT.EXE.
Suppose your terminal type is vt320, and your terminal character-set is
Latin1. Then you "set term type ansi" (or any of the other ANSI types). K95
switches character sets properly, but still treats C1 controls (0x80-0x9F) as
controls, rather than graphics, as they should be in ANSI emulation. This
only happens when switching from a non-ANSI type (and a character set that
uses C1 controls) to an ANSI terminal type. It does not happen if you initial
terminal type is ANSI. Workaround: tell K95 to "set term char transparent"
and all is OK; for example:
set terminal type ansi ; (or scoansi, at386, or avatar)
set terminal character-set transparent
This problem does not affect users of the Dialer unless they use K95 commands
to switch emulations after starting a session. Fixed in 1.1.11.
Emulations that define the Arrow keys to be something other than \Kuparr,
\Kdnarr, ... such as SCOANSI and AT386 terminals that define them as function
keys like \Kansif50, ... do not work as expected for command recall. These
terminal types allow their arrow keys to be programmed by the host, which is
why they must be assigned Kverbs from the User Defined Key list rather than
the regular arrow Kverbs. Workaround: use Ctrl-B and Ctrl-N rather than up
and down arrow, respectively. Fixed in 1.1.11.
In Windows 95, if you choose unsafe screen dimensions (i.e. anything other
than 24x80, 42x80, or 49x80 plus status line, or 25x80, 43x80, or 50x80
without status line) in the Dialer Terminal page, an alert box pops up saying
"heights ... are unsafe". It should say "heights other than ... are unsafe"
(that is, it should say "other than"). Fixed in 1.1.11.
If you tell K95 to "set protocol zmodem" (or ymodem, ymodem-g, or xmodem) and
then give it a SEND command specifying a wildcard, the transfer will fail if
the wildcard matches any directory names, including "." and "..". Thus, for
example, "send *.*" always fails because it always matches ".".
This bug will be fixed in the next release. In the meantime, various
workarounds can be used. Here's one you can use to send all the files from
the current directory, provided (a) it has no subdirectories, and (b) all
files have an extension or filetype (the part after the dot) at least one
character long:
send \?*.\?*
Other workarounds involve MSEND; e.g.:
msend a* b* c* ...
(specifying the first letter of each group of files that you want to send
that does not also match a directory name).
A general workaround with no "but-ifs" is to write a little macro that gets
all the filenames that are not directory names, builds a send list, and then
sends it:
---(cut here)---
declare \&f[100] ; An array is needed
define ZSEND { ; Define a ZSEND macro
local \%i \%j \%n \%f ; Local variables
if not def \%1 end 1 ZSEND what? ; Check for a filespec
asg \%n \ffiles(\%1) ; Expand wildcards if any
asg \%j 0 ; Counter for regular files
for \%i 1 \%n 1 { ; Loop through file list
asg \%f \fnextfile() ; Get next file name
xif not directory \%f { ; If it's not a directory name...
increment \%j ; Count it
asg \&f[\%j] \%f ; and store it in the array.
}
}
clear send-list ; Now make a send list
for \%i 1 \%j 1 { ; Loop through regular file list
echo \%i. \&f[\%i] ; Remove this for silent running
add send-list \&f[\%i] ; Add this file to the send-list
}
echo Sending: ; Remove for silent running
show send-list ; Ditto
send ; Send from the send-list
}
---(cut here)---
Put this in your K95CUSTOM.INI file and then use ZSEND instead of SEND if
you need to use wildcards in your SEND commands when using X-, Y-, or ZMODEM
protocol to upload files. Increase the size of the \&f[] array if you ever
need to send more than 100 files at once.
DEC 7-bit National Replacement Character (NRC) sets are supposed to be used
instead of ASCII when the host requests US ASCII and the terminal is in
Multinational Mode (i.e. the mode which supports 8-bit character sets like
Latin-1 and DEC MCS), but with an NRC defined. This could only happen if
the terminal was set up in National mode using (say) the Swedish NRC, but then
the host switched it into Multinational mode, and then designated ASCII into
G0, G1, G2, and/or G3. Kermit 95 really does use ASCII in this case, whereas
a real international-model DEC VT320 or higher uses the previously-designated
NRC. Fixed in 1.1.11.
ZMODEM crash recovery worked when the FILE COLLISION setting is UPDATE or
OVERWRITE, but not when it is BACKUP, RENAME, DISCARD, etc. It should work
no matter what the FILE COLLISION setting is. Fixed in 1.1.11.
In any VT terminal emulation, VT100 or higher, if a character is written when
the cursor is in the rightmost column, The cursor stays put and a flag is set
to indicate that the next character written should auto-wrap. When the flag
set, and the host sends CR LF ESC [ 79 C, which in effect moves the cursor
down one line, the flag should be turned off but in Kermit 95 it isn't, so the
next character that arrives causes a spurious auto-wrap. Fixed in 1.1.11.
The definition for Shift-F8 contained a typo; an apostrophe instead of a
semicolon where the comment introducer was supposed to be, thus making the
intended comment part of the definition. Fixed in 1.1.11.
An error box would pop up if an attempt was made to print to a nonexistent
or nonready printer; however, there was no way to remove the error box or
continue Kermit 95. This feature was removed in 1.1.11 for Windows 95 and NT,
but remains in the OS/2 version, where it works correctly.
DASHER 200 and 210 terminal types do not support printing. In Kermit 95
1.1.11, the D217 type was added, which does allow printing.
Telnet negotiations received during INPUT/MINPUT, although handled correctly,
could interfere with proper recognition of the target string. Fixed in
1.1.11.
In K95 1.1.10 and earlier, the PBX entry fields would be ignored if the PBX
Checkbox was not selected and the data in those boxes would be lost. Starting
with 1.1.11, the PBX data fields are used and the data is grayed out and
preserved when the checkbox is not selected.
Wyse Graphics characters may be specified as any character above '0' mod 16
instead of just the first 16 characters starting with '0'. Fixed in 1.1.11.
Characters in the range 128-159 were being improperly converted to escape
sequences during transparent-print operations before being passed to the
printer, in VT220/320 and Wyse 370 emulation, when using character sets with
C1 controls. Fixed in 1.1.11.
Kermit 95 was sending <ESC>[?1c when it should have been sending <ESC>[?1;2c
(the ";2" indicates 132-column and attribute support). Fixed in 1.1.11.
The unmodifiable part of the K95 Dialer database, DIALINF.DAT, could not be
installed as a read-only file or on a read-only disk since it was mistakenly
opened in read-write mode. Fixed in 1.1.12.
On a serial connection, when sending a file with a large window and/or packet
size and with SET WIN95 OVERLAPPED-IO ON, transmit buffer overruns could cause
the transfer to fail. Workaround: SET WIN95 OVERLAPPED-IO OFF. Also see bug
number 336. Fixed in 1.1.12.
Bug, fixed in 1.1.12.
If the Dialer was closed before all K95 sessions it had started were
terminated, it could leave temporary files behind. Bug, fixed in 1.1.12.
(See entry 320.)
If the type of the current entry was edited such that it would no longer be
visible based on the active View filters, the Dialer could crash upon exit or
a change in the View filters. Bug, fixed in 1.1.12.
In version 1.1.11 and earlier, SHOW KEY, since it was executed in the Command
screen, displayed the key's value in that context, rather than what it would
send if used in the Terminal screen, which is what most people would expect.
Misleading displays would occur most frequently when a special keyboard was
selected in the Terminal screen, such as EMACS, Hebrew, or Russian. In
version 1.1.12 and later, the Terminal-screen context mapping is displayed,
and it is now also possible to display the mapping from modes that are not
current (see UPDATES.TXT).
If the type of the current entry was edited such that it would no longer be
visible based on the active View filters, the Dialer could crash upon exit or
a change in the View filters. Bug, fixed in 1.1.12.
In version 1.1.11 and earlier, SHOW KEY, since it was executed in the Command
screen, displayed the key's value in that context, rather than what it would
send if used in the Terminal screen, which is what most people would expect.
Misleading displays would occur most frequently when a special keyboard was
selected in the Terminal screen, such as EMACS, Hebrew, or Russian. In
version 1.1.12 and later, the Terminal-screen context mapping is displayed,
and it is now also possible to display the mapping from modes that are not
current (see UPDATES.TXT).
Due to difficulties with i/o redirection in Windows 95 and NT, the REDIRECT
command did not work in all cases. Workarounds were discovered and added in
version 1.1.12, and REDIRECT now works on any connection type, provided the
redirected command is truly redirectable, and in OS/2 too.
The low-level communications device input routines were erroneously discarding
NUL (ASCII 0) characters. This could cause the INPUT command to hang under
certain conditions, and might have had some other undersirable effects. Fixed
in 1.1.12.
Bug, fixed in 1.1.12.
PgUp and PgDn sent the wrong sequences. Definitions were missing for Alt ?+=/
Ctrl-Alt-SP, Alt-DEL. Fixed in 1.1.12.
When sending a file with a window size greater than 1 under conditions of
packet loss, C-Kermit might send packets outside the current window. This
should not cause file corruption, and in many cases (depending on the Kermit
implementation on the receiving end), it won't even cause a noticable problem,
but in other cases it might cause the transfer to stop unnecessarily. Fixed
in 1.1.12.
If you build up a list of files to be sent using ADD SEND-LIST and then
give a SEND command with no operands, it sends the files in the send list.
If you give a MOVE command with no operands, it does the same thing. But
unlike SEND, MOVE should delete each original after sending it successfully;
in Kermit 95 1.1.11 and earlier, it did not. Fixed in 1.1.12.
When the first command-line argument to Kermit 95 is a filename, Kermit 95
should execute the commands from the file, e.g.
k95 filename
But if the filename contains spaces, this didn't work in version 1.1.11 and
earlier, even if the filename is properly quoted to the shell:
k95 "this is a file name"
Fixed in 1.1.12.
Hayes removed an S register (S82) from its newer model Accura modems, and
probably also the others in its high-speed line. "hayes-high-speed" init
string adjusted in version 1.1.12 to account for this.
IF FAILURE did not work after a READ command that was executed before
the READ file was opened. Fixed in 1.1.12.
The 1023-byte limit on function arguments was overly restrictive for Kermit 95.
In version 1.1.12, it was increased to 10238 (approximately 10K).
Certain \f...() functions returned incorrect values, or even caused crashes or
invalid page faults when arguments were missing (especially the first one), or
when they were too long. Also:
- The \fverify() function did not work right when given a starting position
as the third argument.
- \fbreak() did not pay attention to the SET CASE setting
- fverify() did not pay attention to the SET CASE setting
- fverify() would crash if given a huge number as the 3rd argument
- \flpad() and \frpad() erroneously truncated their arguments if they
were longer than the padding length
- \ftrim() and \fltrim() erroneously returned "0" if given an empty argument
- \ttrim() and \fltrim() ignored SET CASE
Also, some of the string functions were inconsistent as to whether they
allowed a starting position -- e.g. \fverify() did, but \fbreak() didn't.
So an optional starting position was added as the 3rd argument to \fbreak()
and \fspan(). These are not documented in manual.
All these problems were fixed in 1.1.12.
"\feval(number%0)", "\fmod(number,0)", or "EVALUATE number % 0" would
crash C-Kermit. Also, dividing by zero would return -1 without issuing any
kind of warning. Now both return -1 and always print a "?Divide by zero"
warning. Version 1.1.12.
The DIAL and REDIAL commands, if given from top level with DIAL CONNECT set
to ON or AUTO, and if they succeed, cause C-Kermit to enter CONNECT mode
automatically. ANSWER should behave the same way. It does as of 1.1.12.
When SET EXIT WARNING is ON, C-Kermit can sometimes warn you about a
connection being open when it couldn't possibly be open. This can happen,
for example, if you SET LINE to a serial device and then HANGUP, but the
serial device is still getting the Carrier Detect signal (e.g. because it is
connected to a modem that never drops CD). However, if C-Kermit has been
told to HANGUP, or to close the connection via SET HOST or SET LINE with no
argument, then it knows the connection is closed and so it should not warn.
This is the case as of version 1.1.12.
Bug, fixed in 1.1.12.
The REMOTE DELETE summary report counted files as being deleted when
they actually were not (e.g. if write-protected). Fixed in 1.1.12.
The CHECK command was not reliable. Fixed in 1.1.12.
SET TRANSMIT ECHO ON didn't work when PARITY was NONE. Fixed in 1.1.12.
In versions 1.1.11 and earlier, Kermit 95 deliberately did not force the modem
to issue "word result codes", but allowed the modem to retain its current
result-code selection (word or digit). If the modem issued digit result
codes, Kermit 95 would try to act on them. However, each make and model of
modem has dozens of incompatible digit result codes. So when dialing a modem
that issues digit result codes, Kermit 95 would sometimes misinterpret the
result and declare that a call failed when it succeeded, or vice versa. Fixed
in 1.1.12 by (a) adding lots of new modem-specific digit results, and (b)
forcing word result codes since these are relatively consistent across most
makes and models. Also in 1.1.12, K95 forces modem commands to be echoed, for
help when debugging problem modem connections with SET DIAL DISPLAY ON.
IF EXIST and IF DIRECTORY were fixed in version 1.1.12 to properly strip
braces from around their arguments, so "if directory {C:\Program Files}", etc,
would work as expected. Previously such commands caused parse errors.
Commands that parsed directory names did not provide the same services as
commands that parsed file names, in particular name completion, and name lists
upon "?". Fixed in 1.1.12.
The READ command was overly picky about "last line lacks terminator". It was
changed in 1.1.12 to allow the last or only line to lack a terminator. This
is also a sneaky way of allowing even binary files to be read into a variable
(all at once) if the READ buffer is set large enough (and the file does not
contain NUL characters).
Since braces are used in function calls to indicate grouping, there is no way
to pass literal braces to the function itself. Solution: Define a variable
containing the string that has braces. Example:
define \%a ab{cd
echo \fsubstring(\%a)
ab{cd
If the string is to start with a leading brace and end with a closing brace,
then double braces must appear around the string (which itself is enclosed in
braces):
define \%a {{{foo}}}
echo \fsubstring(\%a)
{foo}
This also works for any other kind of string:
define \%a {{ab{cd}}
echo \fsubstring(\%a)
ab{cd
The Dialer would open the file DIALINF.DAT in Read/Write mode. This would
prevent it from being installed on a network file server with Read Only and
File Scan privileges. Fixed in 1.1.12.
EMACS mode had incorrect definitions for PgUp/PgDn and was missing definitions
for Alt versions of '?','+','=','/', and DEL and Ctrl-Alt-SP. Fixed in 1.1.12.
The Windows versions of K95 were stripping NULs from the input stream during
file transfers and transparent printing when they shouldn't have been. This
did not cause problems for many users since most file transfer protocol
implementations quote the NUL character, and it is insignificant to most
printers. Fixed in 1.1.12.
The was ignored when used with XYZMODEM file transfers. Fixed in
1.1.12.
Kermit-95 did not always delete startup files depending upon how the program
was terminated and the type of connection that was made. Fixed in 1.1.12.
The SHOW MODEM command did not quote backslashes when displaying modem command
strings. Therefore, new commands could not be created using cut and paste.
The TX Buffer has been increased to 20000 bytes. This is large enough that
all file transfer packets may be written without Windows 95 or NT truncating
the write request. 1.1.12.
If INETD.EXE was executing in the background and not the foreground, K2DC.EXE
would close the incoming socket before K2.EXE was able to access it. OS/2
only. Fixed in 1.1.12.
When auto-print was active, form feeds would result in the current terminal
line being printed twice. Wyse and Televideo terminals would print every
line twice. Fixed in 1.1.12.
Data General emulations use RS instead of ESC as the prefix character for
keyboard command strings. This was not displayed properly by SAVE KEYMAP or
SHOW KEY. Fixed in 1.1.12.
Fixed in 1.1.12. See bug 336.
Hazeltine 1500 emulation was completely broken in 1.1.11 and earlier. Fixed
and tested against real Hazeltine 1500 terminals and applications in 1.1.12.
<ESC> z value string DEL was prematurely terminated by CR.
<ESC> f text CR was not placing text into the status line.
<ESC> F text CR now places text into the Title Bar.
Switching to/from 132-col mode should not have cleared the screen.
Fixed in 1.1.12.
DG WordProcessing, Greek/Math, and LineDrawing character sets did not work.
UNIX mode did not work. And: DG cursor positioning uses pure binary
characters to indicate cursor positioning. Over a NVT TELNET connection we
must beware of the possibility that a NUL will be inserted into the stream
after a CR, an instruction to move to column or line 14. All fixed in 1.1.12.
HPTERM did not handle HP Math/Special and Line-drawing character sets.
These were added in 1.1.12.
All Dialer entries were checked and obsolete ones were removed in 1.1.12.
The ASCII DEL character (127) is supposed to be used for padding in Wyse,
Televideo, and Hazeltine emulations, but instead it was being displayed,
usually as a triangle (or house). Fixed in 1.1.12.
NUL characters embedded in escape sequences were erroneously displayed as
spaces. Fixed in 1.1.12.
Overlapped i/o, which is used by Kermit 95 unless you say otherwise, could
conflict with the SuperLAT Gateway COM port redirectors. Data could also be
lost, corrupted, or duplicated in serial (direct or dialed) connection during
file transfer and scripting (e.g. host mode). Diagnosis: bug. Fixed in
1.1.12.
Again with the backslashes! When parsing an input filename, K95 went to great
lengths to allow single backslashes as directory separators. Thus: "send
\k95\tmp\oofa.txt" worked correctly, so did "send /k95/tmp/oofa.txt".
However, this command did not work if the backslashes were doubled. This was
a bug, fixed in 1.1.13. However, note that you can not mix double and single
backslashes in the same filename.
If you typed Tab or Esc in the middle of input file specification containing
double backslashes, e.g. "send \\k95\\docs\\upda<ESC>", K95 would seem
to hang for a long period of time. When a double backslash was passed to
Windows file lookup service, this would evidently make Windows scan the entire
disk. Eventually it would report that the file was not found. The fix listed
in item 339 also fixes this problem.
The sequence for shutting down K95's various threads could sometimes tickle a
bug in Windows, in which a request to close a handle would never return.
This is worked around in 1.1.13 by adding Yet Another Pause to avoid the race
condition, which in turn requires K95 termination to take a little bit
(milliseconds) longer.
It is possible to have no modems defined in the Dialer. But this could result
in an invalid pointer error when editing entries. Bug, fixed in 1.1.13.
Under certain circumstances, the Dialer could fail to free allocated memory
and so under heavy use, memory or swap space could run low. Bug, fixed in
1.1.13.
If a URL ended with a period because it was at the end of a sentence, and you
Ctrl-clicked it, the period would be included (because period is a valid
character in a URL). This is not exactly a bug, but it makes more sense to
strip a trailing period since real URLs hardly ever have one. Trailing
periods are stripped from Ctrl-clicked-on URLs in 1.1.13.
In 1.1.12, the Windows version of Kermit 95 picks up your preferred editor
from the Windows Registry based on the association for ".txt". In some cases,
"%1" might be listed as a command-line option (when really it is a placeholder
for the filename). In this case, SHOW EDITOR in K95 displays something like:
editor: C:\WINDOWS\NOTEPAD.EXE
options: %1
file: (none)
If you give an EDIT command, e.g. "edit read.me", the editor (NotePad in this
example) is likely to pop up an error box:
Cannot open the %1 D:\K95\READ.ME file.
To work around, put a specific SET EDITOR command in your K95CUSTOM.INI file
to override the value picked up from the registry, e.g.
SET EDITOR C:\WINDOWS\NOTEPAD.EXE
In K95 1.1.13, the "%1" is correctly interpreted as a placeholder for the
filename, and the filename is substituted for it.
Note, however, that if the application has installed itself such that the "%1"
notation is enclosed in quotes with the application filename in the Registry,
this means it is part of the filename, literally. In this case, SHOW EDITOR
in K95 displays something like:
editor: "C:\WINDOWS\SOMENAME.EXE %1"
options: (none)
file: (none)
We have had reports of several applications that install themselves this way,
which is wrong. But there is nothing K95 can do about it, since there is no
way for K95 to distinguish command-line options from pieces of the filename.
For example:
editor: "C:\Windows\Program Files\Some Editor.EXE %1 /X/Y/Z"
options: (none)
file: (none)
If the host sends DO LOCATION but K95 does not have a location string, K95
would say WILL LOCATION, but then not send one. But some hosts would wait
forever for the string to arrive (they shouldn't). Corrected in 1.1.13, in
which K95 says WONT LOCATION if it does not have location string.
If a mouse action was undefined, but that action is used, K95 could become
stuck in marking (text selection) mode. Bug in 1.1.12, fixed in 1.1.13.
Tab and Backtab did not properly wrap. Left/right screen margins and
host-directed printer requests were ignored. Bugs, fixed in 1.1.13.
Wyse emulations did not properly default to autoscroll mode. Wyse 50
emulation failed to update character attributes if DeleteChar is used to
clear an attribute cell. Bugs, fixed in 1.1.13.
If K95 was requested to change its screen dimensions to a legal combination,
but could not do so (e.g. because it was running in fullscreen mode), it
could become stuck. Bug, fixed in 1.1.13.
SET MOUSE ACTIVATE did not allow ON or OFF to be specified. Bug, fixed in
1.1.13. (The synonym, SET TERMINAL MOUSE { ON, OFF }, still worked, and
still does.)
In Windows 95, console windows are required to use DOS (OEM) code pages (also
known as "IBM extended character sets"), whereas GUI windows use Windows
("ANSI") code pages, or "fonts", but not DOS code pages. These two do not
agree, neither in repertoire nor encoding. In 1.1.12 and earlier, if you used
K95's built-in mouse actions to paste from a K95 window to the Clipboard or
vice versa, non-ASCII characters (such as accented or non-Roman letters) were
copied literally, without translation, resulting in "garbage". This is fixed
in 1.1.13 by asking Windows 95 to convert the character set between the two
environments during the copy or paste operation. All translations are done by
Windows, and are exactly the same as you would get between any other console
window (not involving Kermit at all) the Clipboard. Note that in Windows 95,
a code page and a Windows font generally do not have all the same characters,
so some of the translations will necessarily be approximations. There is no
problem in Windows NT or OS/2, because the same character set is used in the
K95 window and the Clipboard.
On ANSI-based terminals (VT100, etc): although transparent printing could
be initiated with an 8-bit CSI, it could only be terminated with a 7-bit
CSI. Thus, 8-bit CSI requests to terminate transparent printing would be
ignored. Bug, fixed in 1.1.13.
The terminal emulator thread did not block during execution of an APC.
So (for example), if K95 received an APC command to change its terminal type,
which requires some time to execute, escape sequences for the new terminal
type might be processed in the terminal emalutor before the switch was
complete. In 1.1.13, the terminal emulator blocks during APC execution.
If K95 was executing an APC command sent by the host which contained an INPUT
command, it would go into an infinite loop until its command stack filled up,
and then it would begin printing spurious error messages. Bug, fixed in
1.1.13 by not allowing additional APCs to be defined if one was already in
progress.
Certain TCP/IP (Winsock) products do not implement the full Winsock
specification -- in particular the ability of the select() call to check for
writeability of a socket, which K95 uses to avoid deadlocks; this select()
call always returns a code saying the socket is not writeable, which results
in TRANSMISSION BLOCKED messages and an overall failure to communicate. This
has been observed with the Iware Connect from Quarterdeck (v1.7) and Cisco
(TGV) TCP/IP products. Quarterdeck reportedly has fixed this bug in their
stack in a subsequent unreleased version available as an Alpha test upon
request.
Workaround: None known, other than putting back the original Microsoft Winsock
stack.
Fixed in 1.1.14 with the addition of SET WIN95 SELECT-BUG {ON,OFF}.
File names with spaces in them can not be read by K95.EXE as part of scripts
generated by the Dialer. This applies to names of key mapping files, names of
login scripts, etc, that are specified in the Dialer. Fixed in 1.1.14.
An infinite loop could be entered if the connection was dropped while the
emulator was in the process of reading a character. Fixed in 1.1.14.
MS-DOS Kermit keycode compatibility mode used the NumLock Off Unshifted
keycodes instead of the NumLock On Unshifted keycodes; this will be changed
in the next patch. However, it should still be noted that MS-DOS Kermit has
one keycode for NumLock Off Shifted and NumLock On Unshifted whereas K95 has
separate keycodes for each. Therefore, it is not possible to assign a
definition to NumLock Off Shifted while in MS-DOS Kermit compatibility mode.
Fixed in 1.1.14.
The OUTPUT command would act like ECHO if there was no connection.
And it would echo text twice if SET INPUT ECHO ON and SET DUPLEX HALF.
Fixed in 1.1.14.
Instead of terminating the current macro or command file, the END command,
when given in a SWITCH statement, just exits from the SWITCH statement.
Workaround: use STOP instead of END if it doesn't make a difference, or else
GOTO a common error exit. Fixed in 1.1.14.
ANSI values produced by Dead key and Alt-Gr key combinations (since 1.1.8)
have been mistranslated in the routines meant to work around Microsoft's
failure to provide support for Dead Keys and Alt-Gr key combinations to 32-bit
console applications. Fixed in 1.1.14.
In the K95 Dialer, the meaning of Options->Dialing->TAPI Modem Dialing is
reversed; when checked it means Kermit dialing, when unchecked TAPI dialing
is used. Workaround: check the opposite box from the one you mean.
If the modem in use is a TAPI modem and if the Registry defines "speed
adjustment" on/off entries, the Dialer provides no way to defeat this; if the
speed adjustment is selected in TAPI and the modem reports (for example)
CONNECT 21600, Kermit tries to change its interface speed to 21600 and fails.
Fixed in 1.1.14.
K-95 says that it is UNIX when it should say either WIN32 or OS2 depending on
the system type. Fixed in 1.1.14.
SET ALARM and IF ALARM, when used with times between midnight and 2:46:40am
will not work due to an internal error. Fixed in 1.1.14.
Fixed in 1.1.14.
Fixed in 1.1.14.
- DG emulation "Load Time of Day" command sequence was being parsed
incorrectly.
- DG and HP Line Drawing characters were not being displayed properly using
OEM character sets (i.e. console code pages).
- ASCII emulations (HP,DG,TVI,Wyse,HZ,...) failed to apply parity to
characters read while parsing Escape command sequences.
- All emulations would fail to display graphics characters properly in
NT due to missing characters in the Lucida Console Unicode font.
- ANSI emulations (SCO,AT386,ANSI-BBS) would not reset the active color
when colors were reset by the host.
- HP emulation was unable to access many functions.
All fixed in 1.1.14.
K95 failed to capture all Ctrl-C events and NT would pass them onto the parent
process (4NT.EXE). 4NT.EXE would receive the event and think that the child
process (K95.EXE) terminated and would attempt to read keyboard input and
update the display which would compete with K95. Fixed in 1.1.14.
When editing an entry in the dialer, if the Login page of the notebook is
not viewed and the printer setting is edited, the contents of the Login
Script Text box will be truncated. Status: Fixed in 1.1.14.
- The "Only connect at this speed" control in the Modems control panel allows
the user to specify whether "Speed Negotiation" of modem to modem data
rates is to be performed by the local modem. This feature was being
misinterpretted by K95 as "Speed Matching". This would prevent K95 from
being able to successfully dial that modem.
- TAPI's "No Flow Control" modem command was not being imported by K95.
- Values for Speed, Flow Control, and other communication line parameters
were not always utilized by TAPI when K95 would assign them.
- K95 rules for importing Flow Control settings from TAPI could result in
a mismatch of the Serial Port and Modem Flow Control settings.
All Fixed in 1.1.14.
K95 had trouble with the high-speed single-character writes used in the
dial routines (Windows only). Fixed in 1.1.14.
Modem Commands in the Dialer should only be reset upon user request. Fixed
in 1.1.14.
If SHOW KEY was terminated with Ctrl-C, future attempts to translate keycode
values into characters, Kverbs, and Macros would fail. Fixed in 1.1.14.
If a Winsock stack did not implement select() to wait for the specified
timeout period before returning (it didn't block), it was possible for K95 to
enter an infinite Spin Loop which would utilize large quantities of high
priority CPU time. This situation cannot occur with the Microsoft Winsock
stacks on Windows 95 or NT. Fixed in 1.1.14.
Telnet negotiation parameters were not properly reset between Telnet sessions
to ports other than the standard telnet port (23). Depending on the results
of the previous sessions negotiations a dead lock situation could occur on
subsequent attempts to connect to the host from the same K95 session. Fixed
in 1.1.14.
Suprasonic, Best Data, AT&T Secure Data, and Keep In-Touch modem definitions
were incorrectly referenced and would utilize the wrong definition. Fixed in
1.1.14.
On Windows 95 or Windows NT 4.0, if Windows Telephony (TAPI) has not been
configured the SETUP.EXE program would crash unexpectedly. Fixed simultaneous
to the release of 1.1.14. However, since 1.1.14 is not shipped on diskettes
(1.1.13 is) users must contact Kermit Tech Support to receive access to the
new SETUP.EXE. Workaround: Use Control Panel->Modems to configure a modem
even if you do not have one.
AT386 and SCOANSI had DEC Kverbs assigned to the keypad and Alt-1 through
Alt-0 keys in the default keymap. Of course this could be worked around
with SET KEY commands. Fixed in 1.1.14.
If SET TERMINAL DEBUG ON was used to view Telnet negotiations to a host that
implements Terminal Type negotiation and the host did not recognize the
current terminal type identification string, K95 would enter an infinite loop.
Fixed in 1.1.14.
FTP PC/TCP is a 16-bit stack which does not specify the size of the parameters
to be used during ioctl() calls as do Winsock and the IBM stacks. Passing the
stack a parameter of the wrong size would corrupt the process stack causing a
runtime exception. Fixed in 1.1.14.
Fixed in 1.1.14
Options assigned for use with the assigned browser were ignored by the
MouseURL kverb used to autostart the browser when a URL is clicked.
Fixed in 1.1.14.
When Netscape Navigator 4.x (Communicator) starts the Telnet client it no
longer passes just the host name and port but also includes the protocol tag
at the beginning. This breaks not just K95 but any other Telnet client
including Microsoft Telnet that ships with Windows 95 and NT.
As of 1.1.14, K95 is now capable of parsing these misformatted hostnames.
SET HOST [:[:][[:] is now accepted. This
form can also be passed on the command line to "-j" or "-J". is ignored.
Not really a bug but now that K95 supports a variety of printer types it is
not possible for DOS COPY to be used to print files so it is now done
internally.
TAPI always puts the modem into no-echo mode so SET DIAL DISPLAY ON did not
show you the commands that were being sent to the modem. Fixed in 1.1.14.
A new version is included with 1.1.14 that is more tolerant of bad or lossy
connections.
BREAK action was incorrectly specified as "return to command mode" rather
than "pass through transparently". Fixed in 1.1.14.
K95 could send packets of certain types that were longer than the maximum
length requested by the other Kermit, particularly S and I (parameter
negotiation) and A (attribute) packets. This could be a problem when the
maximum packet length was less than about 40. Fixed in 1.1.14.
Windows NT uses Unicode to represent on-screen characters. But when a screen
dump was requested by \Kdump the Unicode characters used to represent
non-ASCII text (accented or non-Roman letters, box-drawing chracters, etc)
were not translated back to the local OEM Code Page before being printed.
Fixed in 1.1.14
If HPTERM or VIP7809 are specified as the terminal type in the Dialer, the
resulting script would not contain a SET TERMINAL TYPE command, and VT320
(K95's default terminal type) would be used instead. Fixed in 1.1.14.
"send \%1 \%2", when \%2 was undefined, would improperly act as if the file
had no name and make one up, rather than sending it with its own name.
Fixed in 1.1.14.
This one never quite ends. In 1.1.14 we are very careful never to issue
an exit warning unless we are REALLY SURE that it's needed.
Most notably when a file-transfer failed because there was no Kermit program
on the other end. Fixed in 1.1.14.
If the options field of the default EDITOR (as obtained from the Windows
Registry at startup) was "%s", K95 would fail to substitute the filename for
the "%s". This could result in a failure of the editor to load the file.
Fixed in 1.1.14
Wyse graphics, DEC Special, and DEC Technical could produce the wrong
characters depending on the operating system and local code pages. Fixed
1.1.14.
In 1.1.13 and earlier, large combinations of sliding window and packet
sizes for Kermit transfers required a prior SET BUFFERS command to allocate
a bigger packet window. This is no longer necessary as of 1.1.14.
If you told the host-mode menu to download a wildcard specification, it
would fail. Bug, fixed in 1.1.14.
The 1.1.14 patch might have corrupted the resulting K95.EXE executable and
other files in subtle ways. One effect that was noted in the Windows 95/NT
Intel version was that hostmode logins stopped working. Fixed in 1.1.15.
SET TELNET TERMINAL erroneously resulted in a SET TERMINAL TYPE command
whenever a Telnet terminal type negotiation occurred. Fixed in 1.1.15.
If the remote telnet server did not send the CR-NUL sequence together and
Kermit's window size was 1, Kermit transfers could time out waiting for the
NUL. Fixed in 1.1.15.
AIX and HFT emulation Gray arrow keys did not manipulate the command history
list in the Command screen. Fixed in 1.1.15.
AIX and HFT terminals are not supposed to wrap like a VT terminal. Fixed
in 1.1.15.
At least the following problems showed up in 1.1.14 and 1.1.15:
- Failure to update sections of the screen during terminal emulation;
- Coloration mixups;
- In Wyse emulation, entry into protected mode caused an infinite loop;
- VT scrolling region problems;
- GET would fail after an XMODEM SEND.
These were traced to bugs in the optimizer of the new release of the Microsoft
Visual C/C++ compiler SP2 that was used to build Kermit 95 1.1.15 for Windows.
Rebuilding the program without optimization makes these problems disappear.
K95 1.1.16 was built with SP3, with optimization turned off. Thus the K95.EXE
file is considerably bigger, but the risks are lower (the optimization bugs in
1.1.15 were quite subtle and took a good while to show up and to be properly
diagnosed).
Fixed in the SET TERMINAL command list.
[ Top ] [ Index ] [ K95 Home ] [ Kermit Home ]
Kermit 95 Fixed Bug List / Columbia University / kermit@columbia.edu /
2 April 2002