On Wed, 30 Sep 2020 17:10:25 GMT, Igor Ignatyev <iignat...@openjdk.org> wrote:
>> test/hotspot/jtreg/vmTestbase/nsk/share/jvmti/ArgumentHandler.java line 157: >> >>> 155: return; >>> 156: >>> 157: StringTokenizer st = new StringTokenizer(optionString, " ~,"); >> >> I don't see jvmti_tools.cpp changes as part of this PR. Can you elaborate on >> why this change is needed now. > > `jvmti_tools.cpp` wasn't changed by this PR, `jvmti_tools.cpp` always > accepted ` `, `~` and/or `,` as delimiters[[*]]. > as jtreg splits arguments by space symbol, we can't use ` ` in `-agentlib:`, > so we had to change the delimiter used > there to either `,` or `~`, `nsk/share/jvmti/ArgumentHandler` takes these > options (via `getAgentOptionsString`) and > then parse in `parseOptionString`. but `parseOptionString` was using > `StringTokenizer` w/ default delimiters, ie ` > \t\n\r\f`, and as a result it wasn't splitting options correctly after we > changed the delimiter used in tests' > `-agentlib`. in other words, this change is just fixing an old bug which we > didn't care before as we always used ` ` in > this agent's options. [*]: > https://github.com/openjdk/jdk/blob/master/test/hotspot/jtreg/vmTestbase/nsk/share/jvmti/jvmti_tools.cpp#L224 I guess a better question would be is where is ~ being used in the arguments? >> test/hotspot/jtreg/vmTestbase/nsk/jvmti/scenarios/general_functions/GF08/gf08t003/TestDescription.java >> line 62: >> >>> 60: * nsk.jvmti.scenarios.general_functions.GF08.gf08t003 >>> 61: * jni >>> 62: * Registering JNI native method >> >> The motivation for this change isn't obvious to me, and seems less readable >> than the original code. > > the reason I had to this change is that jtreg doesn't handle `"` in any > special way, so `gf08t003 > nsk.jvmti.scenarios.general_functions.GF08.gf08t003 "Registering JNI native > method" jni` is passed to `gf08t` as > [`gf08t003`, `nsk.jvmti.scenarios.general_functions.GF08.gf08t003`, > `"Registering`, `JNI`, `native`, `method"`, `jni`]. > the way I fixed that was by changing the order of `phrase`, `verboseType`, so > `verboseType` is now 2nd arg, and > `phrase` 3rd and the rest of args joined. the alternative way would be to > teach `gf08t` about `"` and make it join > arguments until it gets an even number of `"` and remove them, this, from my > PoV, would be a bit more confusing. if > you, however, are of strong opinion here, I can go w/ that alternative (or > any other if you have more suggestions) So what code used to handle passing quoted arguments? Is this something that PropertyResolvingWrapper used to do? ------------- PR: https://git.openjdk.java.net/jdk/pull/370
