Crash on OS X?

[malmaud]

My program is sporadically crashing with

* thread #1: tid = 0x9bb299, 0x000000030fe16c48 libmbedcrypto.0.dylib`entropy_gather_internal + 127, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=EXC_I386_GPFLT)
    frame #0: 0x000000030fe16c48 libmbedcrypto.0.dylib`entropy_gather_internal + 127

I am not sure yet if somehow the Julia wrapper is causing this or if it is a genuine upstream bug. I don't have a minimal reproducible example yet, but I'm working on it. I thought in the meanwhile I'd open the issue in case someone has any insights.

[mpg]

Can you (install GNU binutils if needed and) show me the output of gobjdump -d library/libmbedcrypto.dylib | grep -A42 'entropy_gather_internal>:'? I'd like to see what happens at entropy_gather_internal + 127. Also, when you say crashing, can you be a bit more specific: segfault, other sign of memory corruption?

FWIW, we regularly run our tests on OS X (including all tests with ASan and most of them with Valgrind's memcheck) and never got a similar issue so far. Of course it might be that the circumstances that trigger the issue never happen in our tests.

[malmaud]

It is a segfault. Here's the binutils output:

/usr/local/lib> gobjdump -d libmbedcrypto.dylib | grep -A42 'entropy_gather_internal>:'
00000000000167cc <_entropy_gather_internal>:
167cc: 55 push %rbp
167cd: 48 89 e5 mov %rsp,%rbp
167d0: 41 57 push %r15
167d2: 41 56 push %r14
167d4: 41 55 push %r13
167d6: 41 54 push %r12
167d8: 53 push %rbx
167d9: 48 81 ec 98 00 00 00 sub $0x98,%rsp
167e0: 49 89 fd mov %rdi,%r13
167e3: 48 8b 05 16 d8 01 00 mov 0x1d816(%rip),%rax # 34000 <_xtea_test_ct+0x5950>
167ea: 48 8b 00 mov (%rax),%rax
167ed: 48 89 45 d0 mov %rax,-0x30(%rbp)
167f1: 41 8b 8d d8 00 00 00 mov 0xd8(%r13),%ecx
167f8: b8 c0 ff ff ff mov $0xffffffc0,%eax
167fd: 85 c9 test %ecx,%ecx
167ff: 0f 84 90 00 00 00 je 16895 <_entropy_gather_internal+0xc9>
16805: b8 c3 ff ff ff mov $0xffffffc3,%eax
1680a: 0f 8e 85 00 00 00 jle 16895 <_entropy_gather_internal+0xc9>
16810: 49 8d 9d 00 01 00 00 lea 0x100(%r13),%rbx
16817: 45 31 e4 xor %r12d,%r12d
1681a: 4c 8d b5 50 ff ff ff lea -0xb0(%rbp),%r14
16821: 45 31 ff xor %r15d,%r15d
16824: 8b 03 mov (%rbx),%eax
16826: 83 f8 01 cmp $0x1,%eax
16829: 44 0f 44 f8 cmove %eax,%r15d
1682d: 48 c7 85 48 ff ff ff movq $0x0,-0xb8(%rbp)
16834: 00 00 00 00
16838: 48 8b 7b e8 mov -0x18(%rbx),%rdi
1683c: ba 80 00 00 00 mov $0x80,%edx
16841: 4c 89 f6 mov %r14,%rsi
16844: 48 8d 8d 48 ff ff ff lea -0xb8(%rbp),%rcx
1684b: ff 53 e0 callq *-0x20(%rbx)
1684e: 85 c0 test %eax,%eax
16850: 75 43 jne 16895 <_entropy_gather_internal+0xc9>
16852: 48 8b 8d 48 ff ff ff mov -0xb8(%rbp),%rcx
16859: 48 85 c9 test %rcx,%rcx
1685c: 74 1a je 16878 <_entropy_gather_internal+0xac>
1685e: 41 0f b6 f4 movzbl %r12b,%esi
16862: 4c 89 ef mov %r13,%rdi
16865: 4c 89 f2 mov %r14,%rdx
16868: e8 c7 fe ff ff callq 16734 <_entropy_update>
1686d: 48 8b 85 48 ff ff ff mov -0xb8(%rbp),%rax

[malmaud]

Turns out it was an idiosyncratic problem on my end. Thanks for your help.

[mpg]

Glad the problem is resolved. Thanks for keeping us updated.