Set_breakpoint crash

Help!
1 
sb(PI,Line,Id) :- 
   % get src file
   pi_head(PI,Head), clause(Head,_,Ref),
   clause_property(Ref,file(File)),
   % set breakpoint
   set_breakpoint(File,Line,1,Id).

8 ?- sb(mod:somepred/1,1229,I).
[...]received fatal signal 11 (segv)
C-stack trace labeled "crash":
  [0] save_backtrace() at :? [0x7f03031884f1]
  [1] print_c_backtrace() at :? [0x7f03031885ad]
  [2] sigCrashHandler() at :? [0x7f03031886bd]
  [3] __restore_rt() at sigaction.c:? [0x7f0302f3cda0]
  [4] pl_clause_from_source4_va() at pl-srcfile.c:? [0x7f03031499ca]
  [5] PL_next_solution___LD() at :? [0x7f0303191c02]
  [6] query_loop() at :? [0x7f03031d3d88]
  [7] prologToplevel() at :? [0x7f03031d3c1b]
  [8] PL_toplevel() at ??:? [0x7f03031e89eb]
  [9] swipl(+0x10b1) [0x55dfe77370b1]
  [10] __libc_start_main() at ??:? [0x7f0302f27b25]
  [11] swipl(+0x10ee) [0x55dfe77370ee]
Prolog stack:
  [13] system:$clause_from_source/4 [PC=1 in supervisor]
  [12] prolog_breakpoints:set_breakpoint/5 [PC=25 in clause 1]
  [10] sb/3 [PC=30 in clause 1]
  [9] $toplevel:toplevel_call/1 [PC=3 in clause 1]
  [8] $toplevel:stop_backtrace/2 [PC=4 in clause 1]
  [7] $tabling:$wfs_call/2 [PC=17 in clause 1]
  [5] $toplevel:$execute_goal2/3 [PC=31 in clause 1]
  [3] $toplevel:$query_loop/0 [PC=39 in clause 2]
  [2] $toplevel:$runtoplevel/0 [PC=19 in clause 1]
  [1] $toplevel:$toplevel/0 [PC=3 in clause 1]
  [0] system:$c_call_prolog/0 [PC=0 in top query clause]
Running on_halt hooks with status 139
Killing 1530697 with default signal handlers

I am getting a crash when trying to set a breakpoint; this happens with the graphical debugger also; it happens with one prolog source code file, unfortunately I can’t reproduce it with a smaller file, but hopefully the above gives enough information to track down the problem. Running the latest SWI-Prolog (8.5.4-13-g9d2babb2c).

2

No clue. I guess the best is to recompile for debugging. The should get us line numbers. Maybe that is enough. Otherwise we’ll need a debugger …

3

Here is the crash with line numbers

C-stack trace labeled "crash":
  [0] save_backtrace() at /tmp/swipl-devel/src/os/pl-cstack.c:331 [0x7f1362ecc74b]
  [1] print_c_backtrace() at /tmp/swipl-devel/src/os/pl-cstack.c:868 [0x7f1362eccfe3]
  [2] sigCrashHandler() at /tmp/swipl-devel/src/os/pl-cstack.c:908 [0x7f1362ecd15d]
  [3] alt_segv_handler() at /tmp/swipl-devel/src/pl-setup.c:753 [0x7f1362dcefe8]
  [4] __restore_rt() at sigaction.c:? [0x7f1362b28da0]
  [5] pl_clause_from_source4_va() at /tmp/swipl-devel/src/pl-srcfile.c:1819 [0x7f1362e14659]
  [6] PL_next_solution___LD() at /tmp/swipl-devel/src/pl-vmi.c:4585 [0x7f1362d05c9c]
  [7] query_loop() at /tmp/swipl-devel/src/pl-pro.c:147 [0x7f1362d9918e]
  [8] prologToplevel() at /tmp/swipl-devel/src/pl-pro.c:496 [0x7f1362d99e36]
  [9] PL_toplevel() at /tmp/swipl-devel/src/pl-fli.c:4558 [0x7f1362e9295a]
  [10] swipl(+0x11b9) [0x555f0f34d1b9]
  [11] __libc_start_main() at ??:? [0x7f1362b13b25]
  [12] swipl(+0x109e) [0x555f0f34d09e]
Prolog stack:
  [14] system:$clause_from_source/4 [PC=1 in supervisor]
  [13] prolog_breakpoints:set_breakpoint/5 [PC=25 in clause 1]
  [11] sb/3 [PC=30 in clause 1]
  [10] system:<meta-call>/1 [PC=8 in clause -1]
  [9] $toplevel:toplevel_call/1 [PC=3 in clause 1]
  [8] $toplevel:stop_backtrace/2 [PC=4 in clause 1]
  [7] $tabling:$wfs_call/2 [PC=17 in clause 1]
  [5] $toplevel:$execute_goal2/3 [PC=31 in clause 1]
  [3] $toplevel:$query_loop/0 [PC=39 in clause 2]
  [2] $toplevel:$runtoplevel/0 [PC=19 in clause 1]
  [1] $toplevel:$toplevel/0 [PC=3 in clause 1]
  [0] system:$c_call_prolog/0 [PC=0 in top query clause]

Seems like the problem is in pl_clause_from_source4_va() at /tmp/swipl-devel/src/pl-srcfile.c:1819 [0x7f1362e14659]

EDIT: in gdb cl is null which is causing the segfault. No idea why it is null though.

│  1818                                                                                                             │
│  >  1819          if ( cl->source_no == source_no )                                                               │
│     1820          { if ( ln >= (int)cl->line_no )                                                                 │
│     1821            { if ( !c || c->line_no < cl->line_no )       
[...]
Thread 1 "swipl" received signal SIGSEGV, Segmentation fault.
0x00007ffff7c5a659 in pl_clause_from_source4_va (PL__t0=231, PL__ac=4, PL__ctx=0x7fffffffd2a0) at /tmp/swipl-devel/src/pl-srcfile.c:1819
(gdb) print cl
$1 = (Clause) 0x0
(gdb)
4

That should not be possible :frowning: So, most likely something more fundamental is wrong. Any way to reproduce this?

5

I tried to make a small reproducible version but it didn’t reproduce with the smaller source. I will try to bisect the code and see if I can make it reproducible with a small sample. It will most likely be some days.

7

It doesn’t need to be small. The way to find this is first to try enabling addresssanitizer (see CMAKE.md) and see whether we have to do with memory corruption. If not, I typically add a little C function that checks for the error condition and insert that at some places in the source to find where the issue is introduced. Can be tedious :frowning:

8

I’ll report when I have found some more info. Merry Christmas, and have a good time of rest!

9

Found some more information about this bug. In the original prolog source I have a few predicates that are dynamic/1 and thread_local/1:

  • If I comment out the thread_local/1 declaration and leave the predicates as dynamic/1 the bug disappears.
  • I tried to reduce it to a smaller piece of prolog source (with thread_local/1 declarations) but still could not make a small reproducible sample.
  • I’ll keep trying to make a reproducible version.
10

That rings a bell. Note that thread_local always implies dynamic. Pushed a fix. Can’t verify it fixes the crash, but it was surely wrong code and could explain a crash.

1 Like
11

Certainly! The patch fixes it, thanks!

1 Like