On Sat, 21 Mar 2026 13:57:31 GMT, Kurt Miller <[email protected]> wrote:
>> make/common/JdkNativeCompilation.gmk line 240: >> >>> 238: $1_$2_STATIC_LIBRARY := true >>> 239: $1_LIBS += $(LIBZ_LIBS) $(LIBPTHREAD) >>> 240: endif >> >> I see that this is mimicking the line above, but this file isn't meant to be >> used for these kinds of exceptions. I also noticed that there is a >> conditional in java.base/lib/CoreLibraries.gmk handling the static part. It >> would be nice if this could all be handled there instead. > > @erikj79 Thank you for the review. I will reply for this openbsd specific > bit. I've tried to figure out a way to have this only be in > java.base/lib/CoreLibraries.gmk. As far as I can tell the > JdkNativeCompilation.gmk parts are needed for two reasons. > > The `$1_$2_STATIC_LIBRARY := true` part is needed to influence > `ResolveLibPath` so that it can look in the correct directory for the static > lib. Removing it makes `ResolveLibPath` not find the static lib. I suppose > `ResolveLibPath` could be adjusted to prefer the shared library dir and if > the lib is not found, fall back to the static lib dir. Alternatively, is > there a way to have `ResolveLibPath` to obtain the value of > `BUILD_LIBJLI_TYPE`? > > The `$1_LIBS += $(LIBZ_LIBS) $(LIBPTHREAD)` part is needed because the static > lib does not have a way to record additional libraries it depends on like a > shared library has with `NEEDED`. I suppose I could add `$(LIBZ_LIBS) > $(LIBPTHREAD)` in another way that only was for OpenBSD. Let me know your > preference on this. > > It seems like these issues apply equally to aix's need for a static libjli > and gtest too. Ah, I see now what this really is. Please disregard my earlier comment. ------------- PR Review Comment: https://git.openjdk.org/bsd-port/pull/6#discussion_r2988527614
