Benchmark "HashMapBench/HashRemove" segfaults.

[xi2]

Hi. Just trying out Pony and I noticed that on my machine one of the benchmarks segfaults.

$ ./build/release/libponyrt.benchmarks --benchmark_filter='HashMapBench/HashRemove'
Run on (4 X 1245.56 MHz CPU s)
2017-10-27 19:20:01
***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
Benchmark                                     Time           CPU Iterations
---------------------------------------------------------------------------
Segmentation fault

Git bisect tells me that dbd777a introduced this.

My system:

$ ponyc --version
0.20.0-0b2a2d2 [release]
compiled with: llvm 3.9.1 -- cc (Debian 4.9.2-10) 4.9.2
$ uname -a
Linux pc 3.16.0-4-amd64 #1 SMP Debian 3.16.43-2+deb8u5 (2017-09-19) x86_64 GNU/Linux

Hope this is helpful.

[SeanTAllen]

Very helpful. Thank you @xi2.

[SeanTAllen]

@xi2 what commands did you run to build and run the benchmark?

[SeanTAllen]

@Praetonus did you see this?

[xi2]

@SeanTAllen, to build I just ran make on it's own, which built the compiler and the benchmarks. To run the specific benchmark that crashes I ran

./build/release/libponyrt.benchmarks --benchmark_filter='HashMapBench/HashRemove'

As well as the system info I gave earlier, I can say that I am using a debian "Jessie" system (not the current stable, but the one before). LLVM 3.9.1 was built from source with the following build commands

cmake -DCMAKE_BUILD_TYPE=MinSizeRel ..
make -j4

If you cannot reproduce and would like me to try anything else (e.g. different LLVM build options), please ask.

[xi2]

I should also say that running ./build/release/libponyrt.benchmarks alone, the other benchmarks run fine up until the one that crashes. The last one that passes OK before the crash is HashMapBench/HashPutNewResize/1/0.

[SeanTAllen]

I'm on OSX, El Capitan. I get:

➜  ponyc-release-clean ./build/release/libponyrt.benchmarks --benchmark_filter='HashMapBench/HashRemove'                                         git:sean-runtime-backpressure
Run on (8 X 2800 MHz CPU s)
dyld: lazy symbol binding failed: Symbol not found: _clock_gettime
  Referenced from: /usr/local/opt/[email protected]/lib/libc++.1.dylib
  Expected in: /usr/lib/libSystem.B.dylib

dyld: Symbol not found: _clock_gettime
  Referenced from: /usr/local/opt/[email protected]/lib/libc++.1.dylib
  Expected in: /usr/lib/libSystem.B.dylib

[1]    182 trace trap  ./build/release/libponyrt.benchmarks

which is... interesting.

[Praetonus]

Duh, I forgot to take the size of the types into account when replacing new by ponyint_pool_alloc_size. Will make a fix.