Multiple random generators generate the same values

[rick-de-water]

Describe the bug
When using multiple random generators, they use the same seed and thus generate the same values.

Expected behavior
Each random generator uses its own seed and produces different numbers

Reproduction steps

uint8_t a = GENERATE(take(10, random(0, 255)));
uint8_t b = GENERATE(take(10, random(0, 255)));

a and b will always have the same value

Platform information:

• Catch version: v2.9.2

[amitherman95]

I tried to reproduce the bug:
Platform: Windows 10
Compiler: MinGW32
Catch version: 2.9.2

Code:

TEST_CASE("bugTest","[generators]"){
uint8_t a = GENERATE(take(5, random(0, 255)));
uint8_t b = GENERATE(take(5, random(0, 255)));
std::cout << "a=" <<(int)a <<" b="<<(int)b<<std::endl;
}

Output:

a=0 b=0
a=0 b=21
a=0 b=153
a=0 b=228
a=0 b=247
a=21 b=0
a=21 b=21
a=21 b=153
a=21 b=228
a=21 b=247
a=153 b=0
a=153 b=21
a=153 b=153
a=153 b=228
a=153 b=247
a=228 b=0
a=228 b=21
a=228 b=153
a=228 b=228
a=228 b=247
a=247 b=0
a=247 b=21
a=247 b=153
a=247 b=228
a=247 b=247

random() uses the same set over and over
I'll look into that

[horenmar]

Yeah, this is in many ways an intended behaviour -- the idea was to guarantee reproducible results given seed, so that the user can rerun failing test and examine what happened in greater detail.

I see that the linked PR centralizes handling of random generators, I'll have to think about it a bit.

[amitherman95]

Martin, the numbers are still reproducible, but the random generator(m_rand) is kept somewhere else, to avoid reinitialization when a GeneratorTracker decides to clear its child tracker.