> Several instances of function pointers in jdk.crypto.cryptoki are marked with > the dllimport attribute, which should only be applied to symbol declarations, > of which a typedef'd function pointer is not. This would only be useful if a > function pointer defined in the linked dll is desired to be imported, not if > the pointer itself is created locally and used to store a function address. > In addition to being incorrect, at least on the versions of Visual C++ the > JDK supports today, it is also redundant; Typically they are used to avoid an > indirect stub that jumps to the proper entry in the import address table, but > usage of these typedefs involves loading the address of a function and > directly (Usually through GetProcAddress, even in other cases it would simply > be set to the address of a function anyway) assigning it to the pointer > before immediately dispatching when called, which bypasses this procedure > entirely and makes the attribute pointless.
Julian Waters has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains three additional commits since the last revision: - Merge branch 'master' into typedefs - Merge branch 'master' into typedefs - Remove __declspec(dllimport) from pointers ------------- Changes: - all: https://git.openjdk.org/jdk/pull/9353/files - new: https://git.openjdk.org/jdk/pull/9353/files/f6a94f41..f0115500 Webrevs: - full: https://webrevs.openjdk.org/?repo=jdk&pr=9353&range=02 - incr: https://webrevs.openjdk.org/?repo=jdk&pr=9353&range=01-02 Stats: 64346 lines in 627 files changed: 34706 ins; 21440 del; 8200 mod Patch: https://git.openjdk.org/jdk/pull/9353.diff Fetch: git fetch https://git.openjdk.org/jdk pull/9353/head:pull/9353 PR: https://git.openjdk.org/jdk/pull/9353