gEDA: PCB segfault in pinout window

[Karel Kulhavý <clock@twibright.com>]

pcb 20040305 segfaults in pinout window.

I have linked it with -lefence and exported CFLAGS="-O0 -g" before
./configure && make && make install of pcb-20040530.

clock@beton:~$ pcb

  Electric Fence 2.1 Copyright (C) 1987-1998 Bruce Perens.
Segmentation fault (core dumped)

Place a random element and open the pinout window. Do it by selecting screen
menu and then pinout window. It will crash on a segfault.
clock@beton:~$ gdb `which pcb-bin` core
(gdb) bt
#0  0x080c2324 in yy_chk ()
#1  0x459e6ffc in ?? ()
#2  0x080b0ca6 in ClearUndoList ()
#3  0x0804febd in ActionChangeSize ()
#4  0x400db7d8 in ?? ()
#5  0x080c0a6e in CostParameter ()
#6  0x080c133f in CostParameter ()
#7  0x400af61d in ?? ()
#8  0x400502d6 in ?? ()
#9  0x4004ece7 in ?? ()
#10 0x400e3b4a in ?? ()
#11 0x400e4537 in ?? ()
#12 0x400e466c in ?? ()
#13 0x400bc864 in ?? ()
#14 0x400bd3a8 in ?? ()
#15 0x400bd55f in ?? ()
#16 0x400bdab4 in ?? ()
#17 0x0809eaf9 in ParseElementFile ()
#18 0x40273f14 in ?? ()

I don't know what more sophisticated debug features I should turn on to
get some meaningful debugging info. This was even the frame data aren't listed
(the variables the functions were called with).

Should I put it also into the sourceforge buglist? It is annoying for me
because I have to paste all the info into the textarea in the browser which
is very crappy editor compared to vi I am using in mutt. And I have also
to click through several links before getting there - this way I just put
geda-dev into the To: field and mutt automatically expands into the full
e-mail address ;-)

Also I should note that the autoplacement algorithm looks somehow crappy. The
round circles around the polarized capacitors seem to be ignored when scanning
PCB boundaries. Sometimes also resistors and diodes and similar small 2-pin
parts get completely crossed over.

It seems to me that first run yields a crap however second and subsequent runs
yield a reasonable placement without the crossovers. The number of elements
is say 50 and they fit without a problem into 1/5 of the total PCB area (they
end up in upper left corner after the algorithm is run).

Cl<