Fix tests with gmp 6.3 #36006
Fix tests with gmp 6.3 #36006
Conversation
|
That feels like another instance of testing an external component rather than sage itself. I'll have a closer look. |
|
LGTM |
|
The CI fails. Please don't ask me to merge tickets in that case. |
TBH, I always ignore CI, since it's broken most of the time. Look: I tested this PR locally, and I'm preparing the void linux package for 10.1 which needs this. I'm not willing to download, build or test sage-the-monolith myself. When sagemath CI starts working consistently, I will stop ignoring it. I would warmly approve if you had a rule not to merge PRs without a green checkmark, since that would mean breaking CI would be a serious issue. |
|
Documentation preview for this PR (built with commit 0d5506d; changes) is ready! 🎉 |
| ....: 2^(2^63-3) | ||
| ....: except (OverflowError, RuntimeError, FloatingPointError): | ||
| ....: print ('Overflow error') | ||
| ...Overflow error |
There was a problem hiding this comment.
It'd be nice to have some explanation for the ... since it looks like we're already trying to list all of the possible errors and replace them with 'Overflow error'
Otherwise, the CI is happy now...
There was a problem hiding this comment.
IIRC the old gmp used to actually print some stuff in addition to an error; in the previous form of the test this would be catched by the ... of the traceback.
It'd feel better if sagemath would catch all these variations and return always the same exception (say OverflowError), then the doctest would document precisely what happens (unlike here where the doctest is documenting that different things can happen but we don't know which one).
I think getting this merged is still a win. Whoever feels inclined can improve this in a separate PR.
With gmp 6.3, overflow errors are raised as
FloatingPointError