On Tue, 26 Mar 2024 23:01:55 GMT, Magnus Ihse Bursie <i...@openjdk.org> wrote:
>> make/modules/java.desktop/lib/Awt2dLibraries.gmk line 145: >> >>> 143: -delayload:gdi32.dll -delayload:imm32.dll -delayload:ole32.dll >>> \ >>> 144: -delayload:shell32.dll -delayload:shlwapi.dll >>> -delayload:user32.dll \ >>> 145: -delayload:winmm.dll -delayload:winspool.drv, \ >> >> I suppose (presume?) that delayload isn't sensitive to the order ? >> But I do have to ask if you ran all the client automated tests, as well as >> making sure builds work ? >> And surely logical ordering related to dependencies is more important than >> lexical sort order ? > > `/delayload` is not order sensitive, no. > > I'm not sure what you mean by "logical ordering related to dependencies". All > these are dependencies for awt.dll, and there is no intra-dependency relation > between them. > > What you see here is just a single file -- I have gone through the entire > build system to make sure all calls to `SetupJdkLibrary` has been consistent. > Among the things I have standardized is making sure that all libraries are > listed in alphabetical order. It will not affect the final product but will > help us keep the makefile source code clean and consistent. I have run tier 1 + tier 3 on all platforms supported on the Oracle CI system, in addition to the comparison build that shows that the resulting binaries should have only superficial changes. Please advice if you think I need to run any additional testing. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/18486#discussion_r1540213312