RFR: 8282773: Refactor parsing of integer VM options [v3]

[Ioi Lam]

On Thu, 10 Mar 2022 05:33:34 GMT, David Holmes <dholmes at openjdk.org> wrote:

> Looks good.
> 
> The test is rather hard to decipher. I see my requested min +/- 1 and max +/- 1 added, but other boundary cases are not obvious e.g. to test nK for example, we want to check n = (max/K)-1; (max/K); (max/K)+1 and similarly around min. And then the same for M, G, T

I updated the test to check for all edge cases of k/m/g/t suffixes.

> src/hotspot/share/runtime/arguments.cpp line 945:
> 
>> 943:     // point character (.)
>> 944:     // If a string looks like a FP number, it would be parsed by
>> 945:     // set_fp_numeric_flag(). See Arguments::parse_argument().
> 
> Okay but ... can't we create a double from a string that represents a number larger than can be parsed as a long? Anyway I'm just surprised that we wouldn't use the type of the flag to determine whether to use set_numeric_flag or set_fp_numeric_flag, and so always use strod for double flags, regardless of the format of the value.

I filed https://bugs.openjdk.java.net/browse/JDK-8283013 -- "strtod should always be used to parse double VM arguments"

-------------

PR: https://git.openjdk.java.net/jdk/pull/7763