Test arguments printed with 32 bit precision

[mknejp]

The arguments passed to benchmarks are of type int64_t however when they get reported they get truncated to 32 bits. The following is output where the argument keeps increasing beyond int precision and is printed incorrectly (but the benchmark does the right thing).

baseline_push_back/std::vector<int>/67108864/manual_time              33079613 ns   32812500 ns         20
baseline_push_back/std::vector<int>/134217728/manual_time             71565844 ns   71428571 ns          7
baseline_push_back/std::vector<int>/536870912/manual_time            539197109 ns  531250000 ns          1
baseline_push_back/std::vector<int>/1073741824/manual_time           914751775 ns  890625000 ns          1
baseline_push_back/std::vector<int>/-2147483648/manual_time         1806131008 ns 1796875000 ns          1
baseline_push_back/std::vector<int>/0/manual_time                   3629563138 ns 3625000000 ns          1

I think the problem is that

benchmark/src/benchmark_register.cc

Line 185 in edc77a3

instance.name += StrFormat("%d", arg);

should use the %lld format specifier instead.

[dmah42]

Sounds about right. PR welcome :)

…

On Mon, Oct 22, 2018 at 6:47 PM Miro Knejp ***@***.***> wrote: The arguments passed to benchmarks are of type int64_t however when they get reported they get truncated to 32 bits. The following is output where the argument keeps increasing beyond int precision and is printed incorrectly (but the benchmark does the right thing). baseline_push_back/std::vector<int>/67108864/manual_time 33079613 ns 32812500 ns 20 baseline_push_back/std::vector<int>/134217728/manual_time 71565844 ns 71428571 ns 7 baseline_push_back/std::vector<int>/536870912/manual_time 539197109 ns 531250000 ns 1 baseline_push_back/std::vector<int>/1073741824/manual_time 914751775 ns 890625000 ns 1 baseline_push_back/std::vector<int>/-2147483648/manual_time 1806131008 ns 1796875000 ns 1 baseline_push_back/std::vector<int>/0/manual_time 3629563138 ns 3625000000 ns 1 I think the problem is that https://github.com/google/benchmark/blob/edc77a3669026eddc380721d5a3cdccd752b76cb/src/benchmark_register.cc#L185 should use the %lld format specifier instead. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#714>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAfIMhrlXVyZHWJxNYhSkRupkqKEo0Rrks5ungSdgaJpZM4Xz17s> .

[LebedevRI]

Is -Wformat not enabled? Why isn't compiler complaining about that..