On Tue, 14 Feb 2023 16:59:03 GMT, Severin Gehwolf <sgehw...@openjdk.org> wrote:
>> src/java.base/share/native/libjimage/imageFile.hpp line 218: >> >>> 216: // Notes: >>> 217: // - Even though ATTRIBUTE_END is used to mark the end of the >>> attribute stream, >>> 218: // streams will contain non-zero byte values to represent lesser >>> significant bits. >> >> This change is not correct. Maybe it is badly worded but the point is that >> **zeroes can** occur in the stream so testing for zero is insufficient. > > @JimLaskey OK. Perhaps we can be clearer what is meant here exactly. I was > having a hard time deciphering this. It does say `stream will contain zero > byte values to represent lesser significant bits`. **What** are "byte values > to represent lesser significant bits"? To me it sounded like it wanted to say: Since the `ATTRIBUTE_END` isn't a full byte (only 5 bits in a byte), it might be represented as a non-zero value. For example a byte containing `0x07` would equally be `ATTRIBUTE_END` as would a zero byte or a `0x01` byte. `ATTRIBUTE_END` is a `kind` which is encoded with the *most* significant `5` bits. Yet, `ATTRIBUTE_END` isn't a full byte. The least significant `3` bits in the byte represent the `length - 1` - of bytes - in the attribute stream for offset values. That, to me, also would suggest that comparing it to a zero byte value is not sufficient to detect `ATTRIBUTE_END`. ------------- PR: https://git.openjdk.org/jdk/pull/12533
