history

一个通讯程序源码

============= Changes for ECU 3.20 =======================================
last revised Sat Aug 22 15:29:11 EDT 1992

"But in our enthusiasm, we could not resist a radical overhaul of
the system, in which all of its major weaknesses have been
exposed, analyzed, and replaced with new weaknesses."  -- Bruce
Leverett 'Register Allocation in Optimizing Compilers'

1. new commands (procedure and interactive)
   a.  autorz
   b.  dcdwatch

2. dial command reworked
   a. new sub menu for entry add/edit
   b. phone number size increased from 25 to 40 characters
      (some modems limit input string size to 40 characters total
      excluding AT and trailing CR)
   c. Trailing dollar sign '$' appends ~/.ecu/.credit to dialled number:
      .credit has:
          ,,,167167002
      Home dial 1234 sends to modem
          1234,,,167167002

3. help cmd improved: you can stay in a category and ask
   for help with multiple commands and use F1 to return to
   category menu.

4. Major rework inside to reduce but not eliminate the likelihood
the chances ecu will abnormally terminate when the dial command
needs to change to a new line and the new line is not available.
If the new line is busy (in use by another dialout or login),
the request to change is rejected and the old line is retained,
the program survives.  If the new line is unavailable because
of other reasons (read/write denied [chmod] for instance), ecu still
dies. hopefuly this will all be handled by the time of release.

5. [e-mail] added preliminary support for ESIX SVR4
and I added the necessary switches to config.c.  Now, -DISCSVR4
is used for ISCSVR4, -DESIXSVR4 for ESIX and -DSVR4 for "any" SVR4.

6. The "kbdtest" interactive command has been cleaned up enough
for general use.  I hacked it in for debugging early on.
Now it is documented well enough for general use in funckeymap debugging.

7. -P command line switch added for selecting a phone directory other
than the default ~/.ecu/phone.  Use -P fullpath.

8. Any usage error (bad command args) resulted in the death of ecu's
process group.  This provided great incentive to invoke the program
properly, but in this modern day, you are not allowed to discipline
the students :-).

9. An experimental feature has been added to ecuungetty.
If it is compiled with -DECUUNGETTY_CHOWN, it will perform
chown and chmod operations on the line.

When the line is acquired, its ownership is set to the real uid
of the ecu user, its group id to the group id of the process, and
its mode to 622.  This is what happens to xterm ptys.

When the line is released, its ownership is set to uucp
of the ecu user, its group id to uucp, and its mode to 640.
This is what SCO uugetty does when it starts on a line.

You can experiment with this feature by 
    a. placing the line
       LOCAL_CFLAGS = -DECUUNGETTY_CHOWN
       in the file `config.local' in the distribution base directory.
    b. Configure or ./config
    c. remake and reinstall

*You* must decide if this is appropriate for you.  The tradeoff
is between having 666 ttys and having a program which a diligent
hacker could use to change any file in the system to his
ownership. You really should look at bamboozle.c and judge
what level of protection you want to use.  If you are on
a sensitive installation (like you .gov or .mil guys),
you might not want to do it.  Any suggestions for improvement
or additional security.  I thought about pathname checking,
st_dev and st_rdev checking and the like, but nothing simple
and portable has come to mind.

10. From 31840 patch 5,
By the time ECU actually attempts an line open, enough previous
context has been lost that the program cannot recover if the open
fails.  I have always made several checks to try and eliminate
such occurrences.  Now, at least for SCO lines in utmp, the
chances of failure are virtually eliminated.

This patch goes with patch 4 to extend the benefits of ecuungetty
chown.  If ecuungetty chown is enabled and this patch is
installed, we get reduced likelihood of ecu terminating
abnormally when switching comm lines or when selecting the first
line of the session.

   Line access is checked if it is not found in utmp to see if
   an open would fail.

   SCO uugetty does not always put anything in the ut_line field
   of its utmp entry.  However, a de facto convention exists for
   making an inittab id field with the last two characters
   matching the last two characters (the only unique characters)
   of SCO-style comm line names.  This patch causes such a check
   to be made, increasing the likelihood of getting a utmp match

Two new files have been added to this distribution:
   dutmp.c   cc -o dutmp dutmp.c    dumps the utmp; a cheap who -a
   README.3.2v4         a brief blurb related to EAFS file names

ecu 3.19.00 patch 1 - notes 11-15

11. The maximum screen geometry is now 85x80 instead of 43x80.
Columns must be 80.  Lines must be > 20.  Lines
in excess of 85 will be accepted, but ECU will be unable to
maintain a correct screen display if any scrolling occurs.

12. A Nap() bug (used by the procedure nap command as well as
internal functions) could cause the program to hang if napping
was to occur when a receiver or child process terminated.  This
only occurred on systems with -DWORKING_SELECT: SunOS, the SVR4s,
ISC 2.2, for instance.

13.  Remaining program identifiers containing the string "nonansi.*key"
were renamed to use "funckeymap".

14.  On systems with reliable predefines in their native compiler,
(SCO -DM_SYSV, SunOS -Dsun),  Configure guesses at a default answer
for the first question.

15. The End key may now be used to finish adding or editing
a dialing directory entry.

ecu 3.19.00 patch 2 - notes 16 - 18

16. Yet more work on the ecuungetty scheme.  ungetty acquisition
not attempted early enough for "enabled but idle" lines

17. A few patches back, I forgot to note %conn will now return
-1 if no line is open.  Previously it returned 0 in this case.
Now -1 means no line open, 0 means line open but no connection,
>0 is the line file descriptor for the line in an active
connection.

18. The Nap() fixes of the last patch prevented ecurz/ecusz/ecusea
from compiling.  Sorry.

From full release 3.19.05

19.  A preliminary T3000 dialer has been added to the gendial
package.  This modem's future is plastic and the dialer, like
all gendial programs, is just a starting point.  Gendial is
provided and I don't have a lot of time to support the zillions
of conflicting idiosyncratic options folks want.  With that
spew spewed, most of them work fairly well.  Documentation :-) ?

20.  A side effect of getting the T3000 was that *I* got to play
with ECU on a Sun.  This is quite fortunate, since ECU was purportedly
ported many CPU cycles ago.  The Nap() problem has finally been
put to bed... (he said).

From 3.19.05 patch 1 (making 3.19.06) notes 21-24

21. The lCLOCAL() code and its users form the chassis upon
which the DCD watcher is based.  It appeared to work on the
SCO box, but caused problems on SunOS that were difficult to
find.  Thanks to Mark Ashton, n4hgf!ifsbd!cma, this problem
is history.  Many mysterious "hangs" came from a trap
you could fall into on *any* implimentation other than SCO UNIX.
CLOCAL doesn't appear to work the same on SCO 3.2.? as XENIX
and SunOS.  I think [e-mail] will be happy
with this news.

22. cma also discovered that if you have multiple Devices entries
for a line to pick up specific Dialers scripts or dialer programs,
say:

ACU ttya - 2400 t3000-2400
ACU ttya - 9600 t3000-v32

ECU would always pick the first one (disregarding baud rate).
This is fixed now.

Now, if you dial with a directory entry using a /Devices-type or
=devices-type specification for tty, then *that specific* Devices
entry is used to choose the dialer program or Dialers script.  If
no directory entry is involved or the directory entry specifies
"Any" or a specific line, then Devices is searched for the first
entry matching both the current line and the current baud rate.

A note aside: SCO Devices entries appear to me to have a syntax
extension not found in all HDB systems.  The baud rate
field allows a range of baud rates, like:

ACU tty1A - 110-2400 /usr/lib/uucp/dialgHA24

ECU supports this syntax, but your uucico may not.

23. models/funckeymap updated with my dual-purpose Sun/SCO
funckeymap/.Xdefaults configuration.

24. cmostime4 has been included for SCO users (and many hackers)

From 3.19.05 patch 2 (making 3.19.07) notes 25
25.  Dialing directory subform data entry formerly had two
input schemes: on input, you were cajoled into going through
fields one at a time until you reached the end of the form, whereupon
ECU would take away the input form.  On the other hand, when editing
an entry, you were free to cycle through the fields "a la modulo"
in either direction.

Now, both input modes allow liberal cycling.  Pressing the END key
signals your acceptance of the entry.  Pressing ESCape in input
mode aborts the new entry process.  Pressing ESCape in edit mode
behaves exactly like the END key except that any edits made since
the cursor entered the current field are discarded..

From 3.19.10 - notes 26-27

26. more internal cleanup

27. [e-mail] (Robert Lewis) noted that putting the phone number
on log files and screen dumps is a security risk.  Why protect
~/.ecu when the phone numbers are plastered everywhere.  No
more banners.

From 3.19.11 - whoops didn't keep track

28. ecuungetty chown fixes and miscellaneous cleanup

From 3.19.12 - notes 29

29. Added use of ECUFUNCKEY environment variable to override TERM
when loading funckeymap.

3.19.13 internal rev (just superstitious :->)

3.19.14 - notes 30

30. Made changes for SCO 3.2v4/ODT 2.0:
    a. pids are now shorts
    b. getty sets "DIALOUT" on its own

3.19.15 - notes 31