Sinon: 7.2.4 is significant slower than 7.2.3

[Flarna]

Describe the bug
For some reason my testsuite runs significant slower with 7.2.4 compared to 7.2.3.

7.2.3:

real    0m14,750s
user    0m17,557s
sys 0m1,412s
````

7.2.4

real 0m54,398s
user 1m0,317s
sys 0m3,563s
```

To Reproduce
Execute unittests using mocha framework.

Expected behavior
No significant difference degrade between 7.2.4 compared to 7.2.3

Context (please complete the following information):

• Library version: 7.2.4
• Environment: windows and linux
• Example URL:
• Other libraries you are using: mocha

Additional context
It's a closed source project therefore I can't share my tests. If you have any pointers what other info I should provide or what to test let me know.

[mantoni]

My first guess is this change. Can you investigate a little?

[Flarna]

Sure, I will take a closer look during the next days.

[fatso83]

While I think it might be due to upping the lolex version (or some other package): https://github.com/sinonjs/sinon/commit/1431c78cc3ffc0c13bbb8155080ccf14d42af4a0

[Flarna]

Looks like @mantoni was correct. I digged a little deeper and it seems that the change in spy#resetHistory is the main contributor of the degradation in my setup.

I have two solutions in mind here:

1. initialize all spy properties already at creation time (spy#createProxy()) - as a result no extend.nonEnum() is needed in resetHistory() as no new properties will be added there
2. reset only properties which are actually present in resetHistory() instead all of them

I think (1) would be the cleaner approach with an additional benefit to have always the same object layout which helps v8 to create better assembly code.

[Flarna]

I created #1984 with a fix proposal.

[ehmicky]

Sorry for the regression introduced by my PR, thanks for fixing it and also adding some tests!

[mroderick]

@Flarna for completeness, would you mind posting some numbers using the new [email protected]?

[Flarna]

Here are the results of the day, sinon 7.2.5 seems to be not available yet on NPM so I used SHA d0c073c2aeaebab276094600d76d178d52158500 which should be the same:

7.2.3
real    0m15,699s
user    0m18,323s
sys 0m1,731s

real    0m15,932s
user    0m18,427s
sys 0m1,669s

7.2.4
real    0m54,554s
user    1m0,454s
sys 0m3,466s

real    0m54,659s
user    1m0,806s
sys 0m3,292s

sinonjs/sinon#d0c073c2aeaebab276094600d76d178d52158500
real    0m16,752s
user    0m19,516s
sys 0m1,767s

real    0m16,460s
user    0m19,514s
sys 0m1,676s

[mroderick]

sinon 7.2.5 seems to be not available yet on NPM

I've published it now :)

[mroderick]

Thank you for the update. Those results look to be well enough in the same ballpark for be usable 👍