RFR: 8286390: Address possibly lossy conversions in jdk.incubator.foreign moved to java.base [v2]
Jorn Vernee
jvernee at openjdk.java.net
Fri May 13 12:06:45 UTC 2022
On Fri, 13 May 2022 11:17:20 GMT, Jorn Vernee <jvernee at openjdk.org> wrote:
>> src/java.base/share/classes/jdk/internal/foreign/abi/ProgrammableInvoker.java line 215:
>>
>>> 213: for (int i = 0; i < highLevelType.parameterCount(); i++) {
>>> 214: List<Binding> bindings = callingSequence.argumentBindings(i);
>>> 215: argInsertPos += Math.toIntExact(bindings.stream().filter(Binding.VMStore.class::isInstance).count()) + 1;
>>
>> My gut feeling in this case would be that it's a bit strange to use `Math.toIntExact` to do a safe cast when you don't do `Math.addExact` to ensure that the result of the addition will not overflow. I wonder if a simple cast wouldn't be more appropriate - unless you really think that you might have more than Integer.MAX_VALUE bindings (which I doubt :-) ). But that's just my feeling...
>
> Yeah, that's true. I don't really expect this to ever overflow, that would create issues in other cases as well (since the limit on argument count is 255 for instance).
>
> Okay, I'll switch this to a regular cast. (It doesn't matter too much I think because I'm also planning to remove this code: https://github.com/openjdk/jdk/pull/8685)
Fixed.
-------------
PR: https://git.openjdk.java.net/jdk/pull/8697
More information about the core-libs-dev
mailing list