Segfaulting in PGE;Perl6Grammar;Compiler;compile (t/compilers/pge/pge_examples.t)

[wiml]

"make testj" has one segfaulting test on netbsd-ppc. I've narrowed it down to the invocation of p6grammar.'compile' in t/compilers/pge/pge_examples.t. I'm attaching a transcript of a debugger session, but the highlights are:

(gdb) run -D4 -R jit ./poo.pir
Starting program: /home/wiml/perl6/partest/parrot -D4 -R jit ./poo.pir

Program received signal SIGSEGV, Segmentation fault.
0x019ebfdc in Parrot_set_p_i (cur_opcode=0x27098ec, interp=0x1fd10e0)
    at src/ops/set.ops:176
176       VTABLE_set_integer_native(interp, $1, $2);
(gdb) bt  
#0  0x019ebfdc in Parrot_set_p_i (cur_opcode=0x27098ec, interp=0x1fd10e0)
    at src/ops/set.ops:176
#1  0x02dde5c8 in ?? ()
#2  0x01a2e000 in runops_jit (interp=Cannot access memory at address 0x18
) at src/interpreter.c:751
#3  0x01a2d964 in runops_int (interp=0x1e0a000, offset=0)
    at src/interpreter.c:980
...
(gdb) p PDB_backtrace(0x1e0a000)
current instr.: 'parrot;PGE;Perl6Grammar;' pc -1 ((unknown file):-1)
called from Sub 'parrot;PGE;Perl6Grammar;statement' pc 0 (EVAL_5:2)
called from Sub 'parrot;PGE;Perl6Grammar;Compiler;compile' pc 152 (runtime/parrot/library/PGE/Perl6Grammar.pir:135)
called from Sub 'example' pc 0 (./poo.pir:5)
$1 = void

My parrot checkout has some modifications that were necessary to get it to build (see my previous tickets) but I think those changes should only be affecting the exec_save functionality, not jit. The non-JIT runcores work fine.

[wiml]

transcript of the crash

[wiml]

my parrot configuration

[jkeenan]

I get a test failure in t/compilers/pge/pge_examples.t as well. Mine comes during the -j part of make fulltest.

/usr/local/bin/perl t/harness --gc-debug --running-make-test -j --runcore-tests
[snip]
t/compilers/pge/pge_examples...........
#   Failed test 'parse FASTA'#   at t/compilers/pge/pge_examples.t line 55.
# Exited with error code: [SIGNAL 11]
# Received:
# 
# Expected:
# >gi|5524211|gb|AAD44166.1| cytochrome b [Elephas maximus maximus]
# LCLYTHIGRNIYYGSYLYSETWNTGIMLLLITMATAFMGYVLPWGQMSFWGATVITNLFSAIPYIGTNLV
# EWIWGGFSVDKATLNRFFAFHFILPFTMVALAGVHLTFLHETGSNNPLGLTSDSDKIPFHPYYTIKDFLG
# LLILILLLLLLALLSPDMLGDPDNHMPADPLNTPLHIKPEWYFLFAYAILRSVPNKLGGVLALFLSIVIL
# GLMPFLHTSKHRSMMLRPLSQALFWTLTMDLLTLTWIGSQPVEYPYTIIGQMASILYFSIILAFLPIAGX
# IENY
# >poly_a teasing the parser with DNA
# aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
# 
# Looks like you failed 1 test of 2.
 Dubious, test returned 1 (wstat 256, 0x100) 
 Failed 1/2 subtests

This is on Darwin-PPC in the same configuration in which I have built Parrot for > 2 years.

[wiml]

Replying to jkeenan:

I get a test failure in t/compilers/pge/pge_examples.t as well. Mine comes during the -j part of make fulltest.

That looks like the same output I was getting (I trimmed pge_examples.t to discover that it crashed when compiling the FASTA grammar, but I originally noticed it in make testj). So I guess this is probably general to ppc, not specific to netbsd.

[wiml]

I took a stab at narrowing down the revision where this bug originated, but it seems that that particular test has been a problem on ppc for a while. Before revision 32222, it simply hangs (apparently indefinitely) rather than crashing:

• r37660 (trunk) - segfaults in FASTA test in pge_examples.t
• r36845 (release 0.9.1) - doesn't build
• r35856 (release 0.9.0) - segfaults in FASTA test in pge_examples.t
• r34006 (release 0.8.2) - segfaults in FASTA test
• r32858 (release 0.8.1) - segfaults in FASTA test
• r32400 - segfaults in FASTA test
• r32250 - segfaults in FASTA test
• r32230 - segfaults in FASTA test
• r32225 - segfaults in FASTA test
• r32223 - segfaults in FASTA test
• r32222 - segfaults in FASTA test
• r32221 - hangs in pge_examples.t
• r32218 - hangs in pge_examples.t
• r32212 - hangs in pge_examples.t
• r32200 - hangs in pge_examples.t (?)
• r32199 - hangs in pge_examples.t
• r32198 - hangs in pge_examples.t
• r32196 - hangs in pge_examples.t
• r32192 - hangs in pge_examples.t
• r32185 - hangs in pge_examples.t
• r32175 - hangs in pge_examples.t
• r32150 - hangs indefinitely in pge_examples.t
• r32092 (release 0.8.0) - hangs indefinitely in pge_examples.t

On the other hand, very few other tests failed in these revisions, which is a nice indication of the code's stability. (I was doing a 'make testj' to test the JIT behavior specifically.)

[coke]

The current JIT core is deprecated.

[jkeenan]

I now get this on Darwin/PPC:

$ perl t/harness --gc-debug -j t/compilers/pge/pge_examples.t
t/compilers/pge/pge_examples.t .. 2/2 
ok 2 - parse FASTA # TODO Darwin segfault -- TT #479
t/compilers/pge/pge_examples.t .. ok   
All tests successful.

Test Summary Report
-------------------
t/compilers/pge/pge_examples.t (Wstat: 0 Tests: 2 Failed: 0)
  TODO passed:   2
Files=1, Tests=2,  9 wallclock secs ( 0.07 usr  0.04 sys +  3.43 cusr  0.52 csys =  4.06 CPU)
Result: PASS

Is this just an artifact of our ripping out JIT code?

If not, if we can get a report on NetBSD/PPC (as per original post), then we can un-TODO the test and close the ticket.

(errmm, I see Coke already closed the ticket as a wontfix.)

Thank you very much.
kid51