[
https://issues.apache.org/jira/browse/NUMBERS-218?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18128272#comment-18128272
]
Alex Herbert commented on NUMBERS-218:
--------------------------------------
Thanks for the report.
The simple fix would be to raise the max exponent from 30 to 31, and 62 to 63.
I presume this is what you are suggesting. You can raise a PR if you would
like, or I can make the change in master with appropriate tests.
> ArithmeticUtils.pow rejects representable minimum-value powers
> --------------------------------------------------------------
>
> Key: NUMBERS-218
> URL: https://issues.apache.org/jira/browse/NUMBERS-218
> Project: Commons Numbers
> Issue Type: Bug
> Components: core
> Affects Versions: 1.3
> Reporter: Aryan Ganjoo
> Priority: Major
>
> ArithmeticUtils.pow(-2, 31) and ArithmeticUtils.pow(-2L, 63) throw
> ArithmeticException even though their exact results are Integer.MIN_VALUE and
> Long.MIN_VALUE. The API documents overflow only when the result cannot fit.
> Expected: those two calls return the minimum values; neighbouring overflowing
> powers, such as pow(2,31) and pow(-2,32), still throw. Actual: an early
> exponent guard rejects both valid boundary cases.
> A standalone regression harness failed on pristine master
> 9b629bdfd72fc8c499795e32bf12909bf34a1f18 and passed with a local two-line
> guard correction. The new JUnit test and mvn -q -pl commons-numbers-core
> verify passed locally; a BigInteger comparison over a broad base/exponent
> range also found no mismatches after the correction. The fix is local and has
> not yet been submitted as a PR. NUMBERS-151 is a closed performance
> refinement concerning 0^0 and does not report this minimum-value boundary.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
