RFR: 8257800: CompileCommand TypedMethodOptionMatcher::parse_method_pattern() may over consume [v2]
Xin Liu
xliu at openjdk.java.net
Tue Dec 8 21:03:37 UTC 2020
On Tue, 8 Dec 2020 11:45:25 GMT, Tobias Hartmann <thartmann at openjdk.org> wrote:
>>> > Do you mind if I add regression test in #1179 to cover this case?
>>> > Actually, I extend the existing testing framework to support CompileCommand option and it is from the random test of PR-1179 discover this issue.
>>>
>>> Yes, that's fine with me.
>>
>> I change my mind. a new unittest is added.
>>
>> I make this issue separate from https://github.com/openjdk/jdk/pull/1179.
>> Because this bug is from the old form "CompileCommand=option,*,ccstrlist,ControlIntrinsic,+_id". we may remove this form in the future.
>
> I'm seeing failures in testing `test/jdk/com/sun/jdi/EATests.java`:
> CompileCommand: An error occurred during parsing
> Error: Did not specify any method name
> Line: 'dontinline,*::dontinline_*'
>
> Usage: '-XX:CompileCommand=<option>,<method pattern>' - to set boolean option to true
> Usage: '-XX:CompileCommand=<option>,<method pattern>,<value>'
> Use: '-XX:CompileCommand=help' for more information and to list all option.
@TobiHartmann
thank you fore verifying that. it's a corner case. The valid form -XX:CompileCommand=dontinline,*::dontinline_* is treated as an error. the reason is that parse_option_name is a fuzzy matcher. it will match `dontinline` out of 'dontinline_*'.
now I parse_option_name is an exact matcher. original parse_option_name change to match_option_name.
-------------
PR: https://git.openjdk.java.net/jdk/pull/1652
More information about the hotspot-compiler-dev
mailing list