Segfault when opening a layout

[Traumflug]

OS is Ubuntu 11.04 AMD64 with Intel internal graphics. PCB built from scratch with git sources as of today:

    ./autogen.sh && ./configure --disable-doc && make clean && make

Steps to reproduce: open an existing layout with the GUI menu.
Expected behaviour: the layout becomes visible.
Actual behaviour: crash.

Here's the backtrace:

mah@piccard:~/RepRap/pcb$ cd src
mah@piccard:~/RepRap/pcb/src$ ./pcbtest.sh -gdb
GNU gdb (Ubuntu/Linaro 7.2-1ubuntu11) 7.2
[...]
(gdb) run
Starting program: /home/mah/RepRap/pcb/src/pcb --lib-path /home/mah/RepRap/pcb/lib --lib-newlib /home/mah/RepRap/pcb/newlib:/home/mah/RepRap/pcb/lib/pcblib-newlib --element-path /home/mah/RepRap/pcb/lib --font-path /home/mah/RepRap/pcb/src --lib-command-dir /home/mah/RepRap/pcb/lib
[Thread debugging using libthread_db enabled]
[New Thread 0x7fffead5f700 (LWP 5035)]
[New Thread 0x7fffea55e700 (LWP 5036)]
disabling 3D rasterization
[New Thread 0x7fffe0f07700 (LWP 5037)]
[Thread 0x7fffe0f07700 (LWP 5037) exited]
[New Thread 0x7fffe0f07700 (LWP 5038)]
[New Thread 0x7fffe0706700 (LWP 5039)]
[New Thread 0x7fffdff05700 (LWP 5040)]
[Thread 0x7fffdff05700 (LWP 5040) exited]
[Thread 0x7fffead5f700 (LWP 5035) exited]
[Thread 0x7fffe0706700 (LWP 5039) exited]

Program received signal SIGSEGV, Segmentation fault.
0x00000000004631f5 in GetValueEx (val=<value optimized out>,
    units=0x12179ab ",6000,3000,1600:Power,4000,7000,4000,1600:Fat,6000,9000,6000,1600:Skinny,10000,13000,10000,1600", absolute=<value optimized out>,
    extra_units=0x0, default_unit=0x1 <Address 0x1 out of bounds>)
    at misc.c:188
188 if (!scaled && default_unit && *default_unit)
(gdb) bt
#0 0x00000000004631f5 in GetValueEx (val=<value optimized out>,
    units=0x12179ab ",6000,3000,1600:Power,4000,7000,4000,1600:Fat,6000,9000,6000,1600:Skinny,10000,13000,10000,1600", absolute=<value optimized out>,
    extra_units=0x0, default_unit=0x1 <Address 0x1 out of bounds>)
    at misc.c:188
#1 0x000000000046331a in GetNum (s=0x7fffffffb9a8,
    default_unit=<value optimized out>) at misc.c:851
#2 0x000000000046442c in ParseRouteString (
    s=0x12179a7 "3000,6000,3000,1600:Power,4000,7000,4000,1600:Fat,6000,9000,6000,1600:Skinny,10000,13000,10000,1600", routeStyle=<value optimized out>,
    default_unit=0x1 <Address 0x1 out of bounds>) at misc.c:881
#3 0x000000000046ef82 in yyparse () at parse_y.y:657
#4 0x000000000046da2c in Parse (Executable=<value optimized out>,
    Path=0x87fdc0 "",
    Filename=0x1136460 "/home/mah/RepRap/Generation_7_Electronics/OptoEndstop.pcb", Parameter=<value optimized out>) at parse_l.l:285
#5 0x000000000045369a in LoadPCB (
    Filename=0x1136460 "/home/mah/RepRap/Generation_7_Electronics/OptoEndstop.pcb") at file.c:408
#6 0x00000000004287dc in ActionLoadFrom (argc=<value optimized out>,
    argv=<value optimized out>, x=<value optimized out>,
    y=<value optimized out>) at action.c:5884
#7 0x0000000000496fcd in hid_actionv (name=<value optimized out>, argc=2,
    argv=0x7fffffffcde0) at hid/common/actions.c:246
#8 0x0000000000497552 in hid_actionl (name=<value optimized out>)
    at hid/common/actions.c:210

(#9 to #36 removed)

Note: the initial empty layout comes up fine.

[Traumflug (mah-jump-ing)]

Here's another backtrace, made with an unoptimized executable:

#0 0x000000000048091f in GetValueEx (
    val=0x12d2847 "3000,6000,3000,1600:Power,2500,6000,3500,1000:Fat,4000,6000,3500,1000:Skinny,600,2402,1181,600",
    units=0x12d284b ",6000,3000,1600:Power,2500,6000,3500,1000:Fat,4000,6000,3500,1000:Skinny,600,2402,1181,600", absolute=0x7fffffffb79e, extra_units=0x0,
    default_unit=0x1 <Address 0x1 out of bounds>) at misc.c:188
#1 0x00000000004836d5 in GetNum (s=0x7fffffffb7f8,
    default_unit=0x1 <Address 0x1 out of bounds>) at misc.c:851
#2 0x00000000004838af in ParseRouteString (
    s=0x12d2847 "3000,6000,3000,1600:Power,2500,6000,3500,1000:Fat,4000,6000,3500,1000:Skinny,600,2402,1181,600", routeStyle=0xfbd1e8,
    default_unit=0x1 <Address 0x1 out of bounds>) at misc.c:881
#3 0x00000000004924bc in yyparse () at parse_y.y:657

Looks like the string "cmil" sent in parse_y.y:657 doesn't reach it's destination.

[Traumflug (mah-jump-ing)]

After some useful hints on the geda-users list, ths diagnosis is as follows:

parse_y.h, parse_y.c and parse_y.o are not updated automatically after parse_y.y has changed. Likely, a Makefile dependency is missing. After deleting all these and rebuilding, file loading worked fine.

Also likely, the situation is the same for recreating parse_l.c and parse_l.o from parse_l.[hl].

[Traumflug (mah-jump-ing)]

The automake-generated build rules in src/Makefile.in check for existence of some files around parse_y.h instead of rebuilding them in case they're outdated. That's obviously not useful.

Attempts to find where these build rules come from failed, however. One item which I think should be corrected is in the attached patch. The problem remains, though.

[DJ Delorie (djdelorie)]

Did you build with --enable-maintainer-mode ? It should enable by
default if you build out of git, but that might be broken...

[Traumflug (mah-jump-ing)]

Uh, no, didn't know a maintainer mode exists.

[Andrew Poelstra (asp11)]

I think this is broken, DJ. I have to do --enable-maintainer-mode explicitly.

[DJ Delorie (djdelorie)]

When you first run ./configure, watch for the line "checking if maintainer mode is required"
It should be enabled automatically if you're building from git AND building the documentation, otherwise it would default to off.

[Traumflug (mah-jump-ing)]

> if you're building from git AND building the documentation

Well, I'm not building documentation (to save time) and the files not properly refreshed don't belong to documentation.

[Traumflug (mah-jump-ing)]

This was fixed with commit 8b567052b8d6580945e1870debe87792036bf73e:

http://git.geda-project.org/pcb/commit/?id=8b567052b8d6580945e1870debe87792036bf73e

[Chad Parker (parker-charles)]

Marked as released in gEDA project, and six years old. If the problem persists, please file a new bug report.