history

一个通讯程序源码

31. ECUUNGETTY_CHOWN now a semi-permanent feature.  It can
still be turned off by editing config.c or Makefiles.  It
is just too good not to have it in.

3.19.16 - notes 32-37

32. cmostime4 backed out of distribution; if you want it, ask me

33. Direct phone numbers longer than 10 characters got doubled in
one instance. Thanks to felton!rll for pointing this out.

34.  Good news!  -DWORKING_SELECT can be and is now used for
3.2v4; as a side effect, the Nap() wrapper/surrogate is now in
nap.c and all the several Nap() functions in ecusea.c, zcommon.c
and gendial.c have been removed.

Also, if -DWORKING_SELECT is used, Nap() uses ftime() instead of
gettimeofday().  Adding -DUSE_GETTIMEOFDAY will return to the
old usage.  This latter note is for porters, since all currently
supported systems have ftime() and several lack gettimeofday().

35. More work has been done on SCO utmp/Devices automatic tty selection
has been done.  I apologize for the long chain of half-fixes.  Once
again, I think it is fixed.  This time I REALLY believe it :-).
For sure, you should no longer find your /dev/tty chown()'d to uucp.

36.  SCO multiscreen color escape sequences from the remote were
being filtered as if ECU were running on a non-multiscreen.  Now,
if you use ECU on an SCO multiscreen, the full set of ECU escape
sequences, including color, again work as specified in the
manual.  Thanks to [e-mail] for pointing this out.  This bug
crept in when I reorganized keyboard management
(nonansikeys->funckeymap).

37. The ecuungetty chown feature is on permanently.  I added
quite a few security/safety checks to make sure some bug or
another will not make the privileged ecuungetty go wild.

3.19.17 - notes 38

38.  The hz/hzmsec needed by nap.c was not being provided,
calculated by ecusea.c, ecu[rs]z.c and gendial.c.

3.19.18 - notes 39

39.  I had started using <sys/time.h> if 3.2v4.  3.2v4 only has
<sys/time.h> if the TCPRT package is installed .  Basic 3.2v4 does
not have it at all.  So, we'll use <time.h> and <sys/select.h> just
like before.  Thanks for the report go to [e-mail].

3.19.19 - notes 40-44

40.  Added support for 3.2v4 CRTSFL.
Now for SCO, rtscts command arguments:

usage: rtscts [ off | on | no | yes | 0..7 ]

argument | RTSFLOW | CTSFLOW   argument | RTSFLOW | CTSFLOW | CRTSFL
---------+---------+---------  ---------+---------+---------+--------
  off    |   0     |   0         0      |   0     |   0     |
  on     |   0     |   1         1      |   0     |   1     |
  no     |   0     |   0         2      |   1     |   0     |
  yes    |   0     |   1         3      |   1     |   1     |
                                 4      |   0     |   0     |   1

Choice 4 only works on SCO 3.2v4 and ODT 2.0.
As you can see, numeric values are masks.
If the 4 bit is present in the numeric value, it
overrides the lower-order bits:  Specifying 7 as an argument
specifies CRTSFL is to be used if it is supported, otherwise
RTSFLOW and CTSFLOW.

The interactive help and iteractive and procedure command
documentation has been updated.  I added to README.RTSCTS 
some illuminating comments by [e-mail] (Chip Salzenberg).
It was seeing this article that brought CRTSFL off the todo
list and into the code.

41.  Warnings about not using ^\Z in Kermit transfers (^Z suspend)
were added to the 'sk' and 'rk' interactive and procedure commands.

42.  This fix went into 3.19.16 but I forgot to note it: color
sequences sent to ECU on a multiscreen console were not being
passed to the driver.  Now they are..  I only know of color
working on SCO.

43.  Changes to make_dirs() was needed to handle making
directories with pathname components which are symbolic links
(possibly on a read-only file system).

44.  The DCD watcher did not work if you use SCO sio.  I have
been using FAS and forgot a a fact about SCO sio: to toggle
CLOCAL meaningfully, you must open the modem (e.g.  tty2B) rather
than the direct (tty2b) file.  On SCO systems, if the tty filename
has a digit+letter as the last two characters,  ECU will
force the lower case (direct) version to be used.  3.19.19 opens
the upper case (modem) device, but continues to store
and display tty names in their lower case version.

3.19.20 - notes 45-46

45. Somewhere in the lifetime of .19, I sent a copy to a guy.
This revision bump is just to keep from having different
versions called .19 out the door.

46.  On 3.19.17/18, there were various reports of hangs on line 
close, hang up, or HOME prompt when using 3.2v2.  I cannot
reproduce any of these errors.  Please get and send me this 
information for me the next time it happens to you:

Install ecu, but execute the ecu built in the distribution directory.

  a. Is the receiver process active (two ecu processes active)?

  b. The ecu transmitter process will either be the only ecu
     process or the oldest.  Kill -4 the transmitter process and move
     the core file to the distribution directory if you 'cd'ed in ecu.

     You may have to type ^Jstty sane^J on the console it failed on
     (expect no echo) to restore the console to something usable.  This
     may not reinstate your choice of character delete, line kill, etc.

  c. do  'sdb ecu |& tee sdb.log'

  d. do  't', then 'q'

  e. repeat the failing scenario if you found the receiver process
     alive in step a.  If so, repeat a-d above, but this time kill
     the receiver process and do  'sdb ecu |& tee -a sdb.log'

  f. Guess.  (mail me the sdb.log) with a description of
     1) the OS
     2) any non-vendor serial driver
     3) what was going on when it failed (in 100 words or more :-> )

3.19.21 - notes 47-49

47.  The UNIX sio driver may generate a SIGTTOU upon execution of
certain ioctl() functions, namely, XCSETA, XCSETAF, XCSETAW,
TIOCSPGRP, TCSBRK, TCFLSH, and TCXONC.  My ignorance caused the
last few revisions to hang for many upon closing a line or upon
performing miscellaneous operations like sending a break.  All
ecu executables now set job control signals to SIG_IGN.  I hope
this fixes the problem.  (Job control is not supported at this
time.)  

48. config.c has been returned to normal compilation (no -g).
I shipped an alternate config.c "out of turn" after 3.19.20.
Some may have installed it, some not. Therefore, unsharing
this patch replaces any config.c you may have in place.

49. The -F switch was added to the command line arguments:

The new manual section titled "-F path":

    Normally, the $TERM environment variable is used to
    determine the funckeymap entry (keyboard configuration)
    to be used.  Sometimes, the $TERM variable is not
    adequate for identifying your keyboard arrangement.
    The -F switch, like the $ECUFUNCKEY environment
    variable, allows you to use override the funckeymap
    entry used.  For more information, see the section
    titled "Function Key Mapping (Recognition)".

3.19.22 - notes 50-52

50.  This note is only of interest to SCO users.
Previous revisions of SCO were lenient on the user in the choice
of the direct or modem tty device.  You could specify either
type and ecu would force the one it thought appropriate.
Users of ecu under other operating systems have always had to
specify the correct name.

If you have time, I'd love feedback on what part of the manuals
I have missed updating with the new paradigm.

Since there are several types of async drivers and we want
to support them, ecu can no longer make a choice for you.
I did radical surgery on the "SCO tty naming" logic.  As it
stands nowu, on SCO, the tty you specify in a dialing entry is the exact
tty name opened.  The convention of using a lower-case lock name
is still observed.  Also, both senses are searched for in utmp
for conflict determination.  ecuungetty chown()/chmod() both
device file names.

However, your choice of which line type (line placed in tty field
of a directory entry or the tty field of the setup form) depends
upon the driver you are using:

sio  modem   /dev/tty1A
fas  direct  /dev/tty1a

I don't know about others.  Tell me.

Notes:
a. An sio direct line ignores setting of CLOCAL (no DCD watcher)
   and lacks support for hardware flow control.
b. Opening the fas modem line causes various forms of conflict
   resulting in EBUSY being returned by fas for an ecu open or
   a dialer program open.

For now, this is meaningful only on SCO.  In the future, perhaps,
we'll manage an installation-dependent table of what line names
refer to the same device and which are modem, direct, etc.

e.g.,
#line_to_use:modem:direct:others:comment
tty1A:tty1A:tty1a::sio
tty1a:tty1A:tty1a::fas
acu0:acu0:ttya,ttyd0:SunOs

This would re-automate SCO tty line type choices and add it anew
for others.  Then, ecu could for the first time on non-SCO (and
SCO naming to boot) search for all possible lock files and utmp
entries.  What do you think?  Yes, one more damn configuration file
to hack on.

51.  ecusz, ecurz and ecusea are now placed in the ecu library directory
(usually /usr/locsal/lib/ecu).  This is a more appropriate place
since it is meaningful only for ecu to execute them.  Kermit or ckermit
is still searched for through $PATH.  Remove any old ecurz, ecusz
and ecusea executables from your locsl bin directory.

52.  I'm sorry the preformatted manuals you have been getting
were malformed.  The new Makefile in doc failed to supply -mm !

BETA 3.20.01 notes 53-58

53. There is still one installation where /dev/tty is getting
incorrectly chown/chmoded.  There are now three brute-force
tests for an attempt at this (up from two).  The bug
hasn't been fixed, but we ought to know how it happens
this time around.

54. The procedure and interactive command fkmap has been added:

Usage: fkmap                           display current mapping
       fkmap <keyname>                 display single key mapping
       fkmap <keyname> <keylist>       modify a key's mapping
       fkmap -r                        reset to original mapping
       fkmap -s <file>                 append current to file

This command manages the mechanism ECU uses to recognize function
keys when they are entered at the console.  If supplied, the