On Wed, 18 Aug 2021 23:57:38 GMT, Phil Race <p...@openjdk.org> wrote:
> When Ctrl+Space is pressed mac generates a string that contains the single > unicode code point zero. > The fn that converts it from an NSString to a Java String is using > NewStringUTF. > The input to that is a null terminated string which also has zero as the code > point it contains, so > we actually end up with a zero length Java string instead of the intended one > code point in length. > So the fix is to change the way we convert the string. > > There's an existing test CtrlAscii.java which sort of tests some of this but > it isn't asserting that you > get what you expect, its mostly testing you don't get something *unexpected* > .. it will happily pass if > you don't get keyevents. I did not want to change the purpose of that test > for this. > So I wrote a test specific to this Ctrl+Space to verify the fix but also ran > all the standard automated tests too. src/java.desktop/macosx/native/libosxapp/JNIUtilities.m line 55: > 53: jstring jStr = (*env)->NewString(env, buffer, len); > 54: free(buffer); > 55: CHECK_EXCEPTION(); Shouldn't it be better to call CHECK_EXCEPTION_NULL_RETURN() here to return null in case NewString throw exception in which case, I guess jStr will have garbage. Also, it seems https://developer.apple.com/documentation/foundation/nsstring/1408720-getcharacters says the buffer must be large enough...aRange.length*sizeof(unichar) so shouldn't the NSRange be created for (0, len * sizeof(unichar)) ------------- PR: https://git.openjdk.java.net/jdk/pull/5177