⟦bcf92ab0a⟧

TextFile



                ==[ All the bugs that fit, we print ]==

Version 1.0 Old Empire - circa 1978, decompiled from optimized
                                11/70 object 1982-3 (honest!).

Bugs should be reported to ihnp4!druxn!tsf  -  Tom Fisher
Credit for the game belongs to Peter Langston.  I take credit for
any bugs remaining in this version.  If you are using a version
that was modified (in addition to the parametric goodies in empglb.c)
since distribution, try to locate the bug and fix it before reporting it.

The only things considered bugs are those that appear to change the
original intent of the code.  Core dumps, meaningless or inaccurate
messages, and obvious cheating opportunities are considered UN-
intentional.  Enhancements, (such as adding a coastwatch command) are
not listed here.

Items containing "Fix:" are fixed in this release.  The others will be 
fixed later.

1) When bridge head is changed to w, b, or u, the number of supplies
   (o, s, g, b, or c) is multiplied.  Put 50 shells in a bridge
   head.  Change it to a warehouse to get 500 shells.  Also, when
   there is a bridge head on both ends of a span, redesignating
   just one of the bridge heads drops the span.

        Fix: desi.c  -  it loses track of things when hunting around
             for spans attached to a bridge head and other bridge
             heads attached to the spans.  Fix it so it keeps track
             of everything as intended.

2) The command line buffer overflows into the BTU variable.  This is
   used to supply yourself with BTUs whenever in need.

        Fix: prmptrd.c  -  add check for length

3) Repay cheats the lender; loan is considered repaid when nothing
   is paid back. This is a good way to start wars.

        Fix: ledg.c   - calculate interest correctly
             lend.c   - fix max loan size
             coll.c   - fix floating point interest and max
             repa.c   - fix floating point interest and max

4) Spy gets info deep within other country's area instead of limiting
   to just adjacent sectors.
        
        Fix: spy.c  -  initialize chart array completely instead of
                       of leaving some areas subject to whatever was
                       on the stack.

5) If your ship automatically flaks enemy planes in same enemy's harbor,
   location of enemy's current capital position is revealed.

        Fix: fly.c  -  the wu() call should use flakker instead of
                       sect.sct_owned when sending telegram to captain
                       of the target ship.

6) Mov reports 34 civilians when 'v' is done while passing through urban
   sector with 340 civilians.

        Fix: mov.c  -  multiply times tunits in printf

7) When a ship is successfully boarded, it becomes part of the '\0' fleet.
   It should become part of the ' ' (or null) fleet.

        Fix: boar.c  -  set the correct value in the fleet variable

8) Logging in with more than 127 telegrams waiting produces a core dump.

        Fix: empdis.c  -  check for negative nat.nat_tgms

9) Last logged in time for deity and visitor not accurate.

        Fix: empcre.c  -  join deity's sanctuary with 127 civilians

10) Flying anything off a carrier results in immediate splashdown.

        Fix: fly.c  -  initialize bomb weight correctly

11) When bombing ships, fly complains about direction input after bombs
     hit.

        Fix: getstri.c  -  use multiple, rotating buffers to avoid
                           clobbering cp in fly.c

12) When moving more material than a destination sector can hold,  excess
    commodity is destroyed rather than returned as stated.

        Fix: mov.c  -  Fix up fcnt to be accurate if it doesn't all fit.

13) Attack causes core dump (11/70 version) when iot instruction hit.

        Fix: flush.c  -  avoid using supplied output functions. Use
                         standard USG C library instead.

14) Number of characters in telegram file as reported by nation
    command not accurate.

        Fix: nati.c  -  use proper structure for fstat.

15) Compiler errors in empglb.c

        Fix: This got fixed with all the changes to the basic structures
             in empdef.h while porting the game to the Amdahl UTS and
             3b systems.

16) Production sometimes goes negative.  Results in negative civilians,
    gold bars, etc.  Very elusive.  Suspect max127() or update().

17) Signals  SIGHUP, SIGINT, SIGQUIT not handled at all.  This stuff was
    more or less ignored during the decompilation process.  Sorry.

18) Country report heading "time" should be BTU ?

19) Spelling error "capital" in desi.c

        Fix: No change.  Spelling is correct!  Speller needs to read a
             few chapters out of the dictionary.

20) Goofing up add command as deity by entering "XXXXX" when asked for
    country number results in "Welcome back to shell...." and wipes
    out the deity in the nation file.

        Fix: If a non-number is entered when asked for a country number,
             the number is taken to be 0 !!  Add a check, in add.c,
             to see if the result is valid (0 < number < MAXNOC).

21) When country is joined and first two-sector allocation is diagonal,
    the welcome message still claims you have 0,0 and 0,1.

22) cen #1 ?des=x&&  produces "'&es' -- meaningless condition?"

        Fix: In snxtsct(), print out the condition argument instead
             of the working buffer, buf, when an error is detected.
             Now it will print out: "'des=x&&'? -- meaningless condition?".

23) Shells landing in the sea report damage to the sea sector.  Should
    be SPLASH!!

        Fix: This only happened when firing from a fort.  Fixed by
             refreshing the sector buffer (getsect(vx, vy, UP_NONE))
             right before checking if victim sector is sanctuary or sea.

24) Mobility goes negative in sector where crash landing occurs.

25) Nav'ing ships at 2,2 causes your ships at 34,2 to move also ?
    This is in a 64,64 world.  This is caused by [xy]wrap() routines
    mishandling the range check.  Needs new algorithm.

26) World of size other than 64,64 never tested after decompilation.
    There are probably hardcoded 2**6 or 64 references somewhere.  Sorry.

27) World is subject to "continental drift" or "Atlantis Syndrome".
    Sectors get changed to water, mountains, or wilderness in an apparently
    random manner.  In at least one such disaster, damaged sectors were
    identical to the sectors immediately to the west.  It appeared that
    the sector buffer had the x-1,y sector data when the putsect(x,y)
    was done.

        Fix: Not known.  But for now, change update.c to return 0
        immediately when lst_up == 0.  Original routine set lst_up to
        curup, returned 1, which caused putsect to be invoked.  Putsect
        would detect the uninhabited case and set lst_up back to 0.
        This change substantially reduces the number of putsects on
        uninhabited sectors and presumably cuts down the bug exposure.  
        Also change getsect.c to detect impending damage to sector, and
        on this condition, CORE DUMP.
        Change assa.c and atta.c to set lst_up to curup when taking over
        a sector.

28) Telegrams get screwed up.  It appears to be placing data from initial-
    data section in the telegram buffer.  Suspect this happens when more
    than 511 characters are in the message or when there is concurrent
    activity in the telegram file.

29) You get a core dump if you enter numbers that are greater than +127
    or -128 when doing a spy command.  E.g.  spy -156:-14,-25,-23 dumps
    memory.  This is because sargs doesn't check for reasonable values
    on return from atoip().  Suspect that a lot of other commands will
    also behave this way.

30) Mobility in submarines can go positive on the 13th torpedo.

        Fix:  Put check in to make sure that mobility stays negative (-120).

31) "Because of containerization ..." message printed when dropping off
    commodity in the same sector it started from.

        Fix:  Before printing message, check to see if hx,hy are the same
              as sx,sy.  If so, don't print.

32) A memory fault occurs when executing the "upd" command.

        Fix:  Memory fault was in the xytoa() routine.  It was being passed
              only two out of three arguments when update() was trying to 
              send the PLAGUE telegram to the victim country.  Supply the 
              missing third argument and declare char *xytoa().
              All changes are in update.c.

33) "n citizens died of PLAGUE in x,y" bulletin has wrong coordinates.
    This bulletin is sent when some other country does an update of the
    world, causing plague to break out or kill citizens in another country.
    The infected country can't find out where the PLAGUE is.

        Fix:  In update.c correct the third argument to xytoa() in the call
              to wu() which sends the bulletin.

34) Victim ships don't fire back properly.  Freighters will fire back up to 
   127 shells in one blast if the attacker is in the same sector.
   The fleet is never checked in returning fire.  Only ships in the 
   same fleet as the victim ship are supposed to fire back.  
   All ships that are able to fire back, do so. Instead of quitting when they 
   have sunk the attacker, they all take a shot at the bubbles.

        Fix: Make freighters impotent, check the fleet, and stop firing
             back as soon as the attacking ship is sunk.  Send a telegram
             to the defender indicating which ship was sunk.  All changes
             are in fire.c .
   

35) Shells landing in the ocean cause a bulletin to be sent to the Deity.

        Fix:  Put a check to inhibit bulletin in fire.c.

36) Treaties aren't checked properly.  You can fire at a ship and not violate
   an active treaty against such mischief.  If you fire at land, and you
   have a treaty against firing on ships, you will be guilty of violating
   a treaty.

        Fix: Straighten out program flow to perform checks in the right
             order in fire.c.

37) You can fire more than 1 gun from a tender.  This little feature makes
   the tender roughly 5 times more awesome than a battleship !  You can
   get 50% damage on a battleship in one shot.

        Fix: Put check in landgun.c to limit guns to one for a tender.

38) If a fort fires on a ship, and ships in the same fleet return fire, the
   bulletin to the defender says fire was returned on "ship at x,y".  It
   should be "sector x,y".

        Fix: Use "alors" instead of "vlors" to determine whether attacker
             is attacking from land or sea in fire.c.

39) Defending forts don't return fire as documented.  If you fire on a fort,
   that fort should return fire if its able.  Furthermore, if that fort is
   defended by another fort, the defending fort should also clobber the
   attacking item.  Of course the target must be in range, shells and guns
   must be available, etc.  This double defense doesn't appear to be 100%
   functional.

        Fix: Pass float to tfact() instead of integer.  Passing integer
             caused defending forts to have a range of 0 in fire.c.

40) The "short round" calculation for firing shells is in error.  If firing
   north or west, the shell sometimes goes the whole distance instead of
   falling short.

        Fix: Force integer truncation by casting to short in fire.c.

41) Spelling error when flying.  "Barometer @620; seas to rough to fly".

        Fix: fly.c - Fix fmt "...seas too rough to fly".

42) Planes will catch flak from ships that have been sunk (owned by the
    Deity), and also from ships from the same country as the planes.
    The first case happens when entering a sector that has sunken ships
    with shells and guns still aboard.  The second case happens if you
    accidently bomb your own ship.  If there are any of your ships in that
    same sector, they will send up flak instead of regarding it as an
    accident (or training mission).

        Fix: fly.c, flakcheck() - Add check to inhibit flak from same
             country or Deity.

43) You get the message "0 citizens died of PLAGUE...".  Notifying the
    country monarch that nothing happened seems superfluous.

        Fix: No Change.  Although nobody died, the message is notification
             that the plague state of the sector changed.  This is
             valuable information.  The message could read "All citizens
             survived the PLAGUE...", but this would be fixing something
             that ain't broke.

44) It is possible to bomb a Deity-owned ship.  Ships owned by the Deity
    should not be detectable.

        Fix: fly.c - When bombing, a check is made to see if the victim ship
             is under the plane.  Add a check to disallow bombing a ship
             owned by the Deity.

45) No information is given about ships when flying over and "viewing"
    sector containing ships.

        Fix: fly.c - Activate previously unreachable code segment.  If ship
             is owned by same country as plane, ship details are given.
             If ship is foreign, two things happen: 1) the ship type and ship
             number (but not country) are given to the pilot, and 2)
             a telegram is sent to flagship's owner "2 country #5 planes
             buzzed ship #123".
             Now you can find out a submarine's ship number one more way
             (besides sonar and look), but not its country.  And now you
             have one more chance for a successful landing when you try to
             land on a carrier and forget the carrier's ship number.

46) When crash landing planes with bombs into ships or sectors, sometimes
    no damage is done.

        Fix: fly.c - In the case where all the planes blew up, there was a
             missing putsect() and putship() to update the sector or ship.
             Thus, the sector or ship would not be updated with all the damage
             done by the exploding bombs.
             With this fixed, Kamikazi attacks are much more effective.

47) In mtch.c, the stmtch() function has a built-in assumption that
    pointers are 4 bytes long.

        Fix: Use a variety of casting operations to get the pointer
             arithmetic to come up with the right answer.  Even lint
             buys it, honest !!

48) If sectors are granted to the Deity, and then the Deity grants those
    sectors to a country, the timestamp of a sector is left at zero.
    The sector will never be updated after that.  The granting operation
    should set the timestamp to curup.

        Fix: In gran.c set sect.sct_lstup = curup before doing the
             doing the putsect.

49) Since the fixes to the "fire" command, shelling land owned by the
    Deity doesn't do any damage to that sector.  

        Fix: In fire.c, a check is made to see if the victim sector
             is owned by the Deity or the attacking country.  In that
             case, no check is made to see if a defending fort can fire
             back -- it just returns.  The fix is to write the victim
             sector out (possibly with damage from the shell) before
             returning.

50) This code doesn't work on a Amdahl UTS UNIX system.

        Fix:  Most of the problems stem from the fact that in the original
             code, characters were used to hold negative values.  This
             doesn't work out on the UTS system since characters are
             unsigned.   Change all the basic structures (gulp!) in empdef.h
             to replace all characters with possible negative values with
             shorts.  Change all the subroutines, commands, and empglb.c
             that have knowledge of the position of these characters, --
             and run lint on the whole thing to get rid of all complaints.
             Whew!  
             Also - get rid of the superet() function that happens to
             have knowledge of the C stack frame layout.  Change all functions
             that called superet() to achieve the functionality a different
             way.

51) The code that was ported to the Amdahl UTS system doesn't work on the 3b.
    There are two problems:  1) chan.c won't compile because it can't find
    the "N_NAME_CHANGE" symbol, and 2) it won't load because it can't find
    the _doprnt() symbol.

        Fix: Change the symbol "N_NAME_CHANGE" in chan.c to "N_NAME_CHNG"
             to match the header file.  The 3b was looking at more characters.

             The 3b libc.a doesn't have the portable version of _doprnt().
             It uses the optimized version (assembly language) called
             __doprnt().  This all stems from a highly questionable choice
             made during decompilation & integration testing -- to use
             the _doprnt() routine directly, rather than indirectly through
             printf or fprintf.  This short-cut backfired on the 3b.   Sorry.
             So OK we have to bite the big bullet here:
             Change all pr(), prfmt(), and pr(fmt()) calls to printf().
             Change all fmt() calls to sprintf(fmtbuf,...) calls, and
             position the sprintf call just before the function needing
             the formatted goodies in fmtbuf.  Change all function calls
             that used fmt() in the calling parameters to use fmtbuf
             instead.  Get rid of explicit fmtlen references in tele.c.
             Comment out most of pr.c, leaving only the global fmtbuf[].
             To get performance back to what it was with the original
             functions, get rid of setbuf(stdout,(char *)NULL) in empdis.c, 
             and insert a fflush(stdout) in prmptrd.c, chkok.c, and cleared.c .
             So now we're a little more portable, I think.  Neat.
DataMuseum.dk

DKUUG/EUUG Conference tapes

⟦bcf92ab0a⟧ TextFile

Derivation

TextFile
