Fix automatic retries of failed requests

[Changaco]

Unless I'm missing something, the "less than" comparison here means that the request is retried n - 1 times instead of n times, so setting max_network_retries to 1 doesn't work and setting it to 2 only retries once.

[brandur-stripe]

@ob-stripe You're more familiar with the code here, so will leave to you. It does look like something is off — we increment the number of retries on the first request rather than waiting for an actual retry. The test looks right, but I think it's measuring retries including the original request, which is why it works.

[ob-stripe]

Ah yeah, looks like num_retries is incremented before the first request.

I think a slightly better fix would be to keep the comparison as is and move the num_retries increment from here to here, because the first request is not a "retry" and should not be counted as such. It would also be slightly more consistent with the stripe-ruby implementation (see here).

[brandur-stripe]

I think a slightly better fix would be to keep the comparison as is and move the num_retries increment from here to here, because the first request is not a "retry" and should not be counted as such. It would also be slightly more consistent with the stripe-ruby implementation (see here).

+1. Going off variable names, the code is pretty confusing at right now which is probably how this bug came in in the first place.

[Changaco]

I think a slightly better fix would be to keep the comparison as is and move the num_retries increment

Done.

[Changaco]

Ping.

[ob-stripe]

Sorry for the delay @Changaco! Thanks again for the contribution.

[ob-stripe]

Released as 2.35.1.