Segfault in __udivmodti4 quickcheck when remainder is requested

[est31]

PR #133 adds quickcheck tests for __udivmodti4 with the code that does checking with remainder enabled commented out. I commented it out as I got a segfault.

[est31]

The segfault happens at:

0x00007ffff55fc22e in __udivmodti4 (a=11837846198047054611, b=10468428215946443262, rem=0x7ffff67fc388)
    at compiler-rt/lib/builtins/udivmodti4.c:47
47                      *rem = n.s.low % d.s.low;

[japaric]

According to this Travis build, this appears to only happens on MIPS.

EDIT: Actually, this might be a different error.

[est31]

@japaric as you can see from this travis log, it failed on multiple 64 bit platforms (x86_64, aarch64, mips64, powerpc64, ...): https://travis-ci.org/rust-lang-nursery/compiler-builtins/builds/198218713

[japaric]

From #155, I'm seeing errors in the following unit tests only on 32-bit MIPS

• divti3, SIGSEGV
• modti3, fails an assert_eq where both LHS and RHS are the same
• muloti4, "attempt to subtract with overflow" src/int/mul.rs:58
• udivmodti4, SIGSEV
• udivt3, wrong answer
  • expected: ((1844674407370955161400, 18002795302052819272), 11689864710784774277)
  • got: ((1844674407370955161400, 18002795302052819272), 11689864690002821119)
  • format: (inputs, output)
• umodti3, "attempt to subtract with overflow" src/int/udiv.rs:278

[japaric]

@est31 AFAICS, all the errors in the build you have linked seem to be due to "missing symbol" which indicates a bad cache.

[est31]

@japaric ah you are right. I think I remember though reproducing the issue on my native platform (x86_64). Maybe you can try to comment out this code and try whether you can reproduce?

[japaric]

Maybe you can try to comment out this code and try whether you can reproduce?

I can locally repro on x86_64. The weird thing is that we are not hitting this bug on x86_64 and other arches in #155.