Backport PR #27631 on branch 0.25.x (BUG: raise when wrong level name is passed to "unstack") #27640
Conversation
…sed to "unstack"
gfyoung
added
Error Reporting
|
@jreback @TomAugspurger @toobaz: Is there a reason for back-porting this? The bug that was being addressed was more than two years old. |
|
The only reason I can think of is "this is a very simple patch". But definitely not urgent. |
|
(Oh, and: over time, backporting patches might be easier if we do it routinely, so the code is more similar. But I have no idea of whether we have any general rule) |
|
Yeah, simple bug fixes are (IMO) fine to backport. It's relatively painless
these days.
…On Tue, Jul 30, 2019 at 5:57 AM Pietro Battiston ***@***.***> wrote:
(Oh, and: over time, backporting patches might be easier if we do it
routinely, so the code is more similar. But I have no idea of whether we
have any general rule)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#27640?email_source=notifications&email_token=AAKAOIUPDATECJJK7YF2IH3QCANALA5CNFSM4IHULDA2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD3DS7SA#issuecomment-516370376>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAKAOIXN2K2QQYIV2L7EX43QCANALANCNFSM4IHULDAQ>
.
|
|
yeah this was fine (and note already was for 0.25.1) |
But note that we should still be careful, and not just backport any bug fix. It happens regularly that a bug fix causes a new bug. |
I agree, but in general should this be a stopper? I mean: do we consider regressions in 0.25.* more problematic than regressions in 1.*? |
Yes, we do (IMO). People should more or less be able to carelessly update 0.25.0 to 0.25.1 while updating a x.y to x.y+1 always should be done carefully (as there can always be API breakages anyway). |
Right, I missed that, I had just in mind the comparison between 0.25.0->0.25.1 and, say, 1.0.0->1.0.1. |
Backport PR #27631: BUG: raise when wrong level name is passed to "unstack"