relax timeouts in terminal console and tests #3817
Conversation
and use kernel_info request to check instead of executions
should decrease false failures on slow VMs (travis)
|
+1 |
|
looks ok to me, but I'm in general opposed to having any kind of magic number naked in source code. With the exception of 0 and 1, it's preferable to have named constants. I don't know what (if any) the relationship between the numbers currently being used, or if they're just ad-hoc, and if it's at all possible to make them all one so that we could just set them in one place as opposed to search-replacing in a few places) |
|
good point - made it Configurable, since that's what those are for. |
as previous commit for test_console
|
and centralized the timeout values in the tests. There is a chance this PR will resolve false-positive failures in Travis, though I have said that before. |
There is a probability distribution that this PR get merged after time t > 0. I would suspect that the cumulative distribution tends to 1 when t go toward infinity. |
|
@Carreau: I think you meant to say there's not a probability distribution that this PR doesn't get merged as t goes to infinity (testing and checking this updated PR now) |
|
@ivanov I didn't wanted not to say it. Don't be angry. |
|
So angry. |
Relax timeouts in terminal console and tests avoids failures on slow kernels (especially on test VMs) uses kernel_info to check if the kernel is alive, rather than executions.
|
Those failure are annoying me a lot. |
| # heart failed | ||
| if timeout is not None and (time.time() - tic) > timeout: | ||
| return False | ||
| return False | ||
| return True |
There was a problem hiding this comment.
there's no way to get to this line, if I'm following the logic properly. Is that right? (if so, remove it?)
|
oops, race condition with @Carreau, who's trigger happy with the merge button ;) |
Relax timeouts in terminal console and tests avoids failures on slow kernels (especially on test VMs) uses kernel_info to check if the kernel is alive, rather than executions.
avoids failures on slow kernels (especially on test VMs)
uses kernel_info to check if the kernel is alive, rather than executions.