Relaxed stride checking fixup

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Relaxed stride checking fixup

Charles R Harris
Hi All,

Currently numpy master has a bogus stride that will cause an error when downstream projects misuse it. That is done in order to help smoke out errors. Previously that bogus stride has been fixed up for releases, but that requires a special patch to be applied after each version branch is made. At this point I'd like to pick one or the other option and make the development and release branches the same in this regard. The question is: which option to choose? Keeping the fixup in master will remove some code and keep things simple, while not fixing up the release will possibly lead to more folks finding errors. At this point in time I am favoring applying the fixup in master.

Thoughts?

Chuck

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Relaxed stride checking fixup

ralfgommers


On Thu, Apr 20, 2017 at 6:28 AM, Charles R Harris <[hidden email]> wrote:
Hi All,

Currently numpy master has a bogus stride that will cause an error when downstream projects misuse it. That is done in order to help smoke out errors. Previously that bogus stride has been fixed up for releases, but that requires a special patch to be applied after each version branch is made. At this point I'd like to pick one or the other option and make the development and release branches the same in this regard. The question is: which option to choose? Keeping the fixup in master will remove some code and keep things simple, while not fixing up the release will possibly lead to more folks finding errors. At this point in time I am favoring applying the fixup in master.

Thoughts?

If we have to pick then keeping the fixup sounds reasonable. Would there be value in making the behavior configurable at compile time? If there are more such things and they'd be behind a __NUMPY_DEBUG__ switch, then people may want to test that in their own CI.

Ralf
 


_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Relaxed stride checking fixup

Charles R Harris


On Thu, Apr 20, 2017 at 4:21 AM, Ralf Gommers <[hidden email]> wrote:


On Thu, Apr 20, 2017 at 6:28 AM, Charles R Harris <[hidden email]> wrote:
Hi All,

Currently numpy master has a bogus stride that will cause an error when downstream projects misuse it. That is done in order to help smoke out errors. Previously that bogus stride has been fixed up for releases, but that requires a special patch to be applied after each version branch is made. At this point I'd like to pick one or the other option and make the development and release branches the same in this regard. The question is: which option to choose? Keeping the fixup in master will remove some code and keep things simple, while not fixing up the release will possibly lead to more folks finding errors. At this point in time I am favoring applying the fixup in master.

Thoughts?

If we have to pick then keeping the fixup sounds reasonable. Would there be value in making the behavior configurable at compile time? If there are more such things and they'd be behind a __NUMPY_DEBUG__ switch, then people may want to test that in their own CI.

Interesting thought. I wonder what else might be a good candidate for such a switch?

Chuck


_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Loading...