Allow interrupting input() on Windows, as part of effort to make pdb interruptible #498
Conversation
|
Not sure what's up with failure on Travis, unless I did something wrong I also get them on |
itamarst
changed the title
|
@Carreau @blink1073 any chance someone could look at this? Lack of |
ipykernel/kernelbase.py
Outdated
| # doing a blocking recv() means stdin reads are uninterruptible | ||
| # on Windows: | ||
| if (ident, reply) == (None, None): | ||
| time.sleep(0.001) |
There was a problem hiding this comment.
I'm wary of spin loops like this; since it is a socket we can use select().
There was a problem hiding this comment.
Unfortunately select() is also uninterruptible by KeyboardInterrupt on Windows. I decided to try using it anyway, because that way at least the loop will be more responsive while needing to wake up less. But sadly as far as I can tell this needs a loop of some sort if it's going to be uninterruptible 😞
There was a problem hiding this comment.
Shucks, thanks for looking into it. I'm not sure what happened on the last travis build, I kicked it.
There was a problem hiding this comment.
I tried matplotlib 3.2.1 on my Linux machine and can't reproduce those issues. But on newer version of Python.
There was a problem hiding this comment.
IPython 7.14.0 was released 4 days ago, that might be it.
There was a problem hiding this comment.
This might be relevant: https://github.com/ipython/ipython/pull/12184/files
There was a problem hiding this comment.
I'm using same version of matplotlib as Travis, though. With updated ipython still can't reproduce... Can you try rerunning tests on master? At least that way can tell if it's in any way related to this PR.
There was a problem hiding this comment.
There was a problem hiding this comment.
...and that failed in the same way
There was a problem hiding this comment.
one thing to try is pinning the ipython version to 7.13.0
This is part of the solution to ipython/ipython#10516.