Running test on mac-os

[berserkr]

I am getting segfaults with the current master branch on mac OS. Somehow the function ptr is not reference properly...

Platform:

Apple LLVM version 8.1.0 (clang-802.0.42)
Target: x86_64-apple-darwin16.6.0
Thread model: posix
InstalledDir: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin

Here is some debugging information:

There is a running process, kill it and restart?: [Y/n] yes
Process 25413 exited with status = 9 (0x00000009)
Process 25418 launched: './testnoham' (x86_64)
Running tests...
test_ntruprime_inv_int ✓
test_ntruprime_inv_poly ✓
testnoham was compiled with optimization - stepping may behave oddly; variables may not be available.
Process 25418 stopped

• thread test_key() stack date currupted #1, queue = 'com.apple.main-thread', stop reason = breakpoint 1.1
  frame #0: 0x000000010000495a testnoham`test_mult_int at test_poly.c:82 [opt]
  79 NtruIntPoly a1 = {11, {-1, 1, 1, 0, -1, 0, 1, 0, 0, 1, -1}};
  80 NtruIntPoly b1 = {11, {14, 11, 26, 24, 14, 16, 30, 7, 25, 6, 19}};
  81 NtruIntPoly c1;
  -> 82 ntru_mult_int(&a1, &b1, &c1, 32-1);
  83 NtruIntPoly c1_exp = {11, {3, 25, -10, 21, 10, 7, 6, 7, 5, 29, -7}};
  84 valid &= equals_poly_mod(&c1_exp, &c1, 32);
  85
  (lldb) print ntru_mult_int
  (uint8_t (*)(NtruIntPoly *, NtruIntPoly *, NtruIntPoly *, uint16_t)) $0 = 0x0000000000000000
  (lldb) print &ntru_mult_int
  (uint8_t (**)(NtruIntPoly *, NtruIntPoly *, NtruIntPoly *, uint16_t)) $1 = 0x000000010000a328
  (lldb) print ntru_mult_int
  (uint8_t ()(NtruIntPoly *, NtruIntPoly *, NtruIntPoly *, uint16_t)) $2 = 0x0000000000000000
  (lldb) step
  Process 25418 stopped
• thread test_key() stack date currupted #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
  frame #0: 0x0000000000000000
  error: memory read failed for 0x0
  (lldb) bt
• thread test_key() stack date currupted #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
  • frame #0: 0x0000000000000000
    frame test_key() stack date currupted #1: 0x000000010000497a testnohamtest_mult_int at test_poly.c:82 [opt] frame #2: 0x00000001000059b1 testnohamtest_poly at test_poly.c:313 [opt]
    frame MGF function incorect realization #3: 0x00000001000044f5 testnohammain(argc=<unavailable>, argv=<unavailable>) at test.c:16 [opt] frame #4: 0x00007fffad4de235 libdyld.dylibstart + 1
    frame Possible memory leak in ntru_invert() #5: 0x00007fffad4de235 libdyld.dylib`start + 1
    (lldb)

It is as if the function ptr are not allocated: ntru_mult_int

I know for sure, it is hitting:

#elif __SSSE3__ ntru_mult_int = &ntru_mult_int_sse; ntru_mult_tern = ntru_mult_tern_sse; ntru_to_arr = ntru_to_arr_sse; ntru_mod_mask = ntru_mod_sse;

in poly.c, however, not sure what is going on :( Any ideas?

[pschlump]

I am seeing this issue also.
I have tested on Linux (ubuntu 16.04) and it is working. On Mac OS I have tried it with both
the clang compiler from Apple (XCode) and with gcc 7.1.0 and I get the same error.

[pschlump]

The problem is global variables that are re-declared in multiple places instead of declared once and then "extern" in other places. Specifically there are 5 function pointers in src/poly.h that need to be
extern in the .h file and then declared in src/poly.c.

I will have a pull request within a day - as soon as I have tested this across all the different architectures. The pull request will also allow for compilation with llvm (clang) on mac.

[berserkr]

Working for me now after applying your changes @pschlump

[pschlump]

The reason for using the ifdef(s) in the code instead of just disabling SSE/AVX2 is because the problem is compile dependent - not mac-os dependent. I have not determined how to get the
clang compiler to generate/use SSE/AVX2 - the gcc compiler (not the fake gcc that ships with
apple's dev tools, but is actually clang) will compile the code correctly. There is no reason to
punish the people using gcc with bad performance when only the clang folks have the problem.

If somebody knows how to get SSE/AVX2 to work in clang please add a note to this and I will re-build
the pull request with that in it.

[mouse07410]

#47 fixes this and #46. Tested with GCC-10 and Clang-10.