On Mon, 30 Oct 2023 10:54:48 GMT, suchismith1993 <d...@openjdk.org> wrote:
> 1. Adding required compiler flags. > 2. Adding required symbols. > > JBS-ISSUE : [JDK-8317799](https://bugs.openjdk.org/browse/JDK-8317799) > Problem: There is syslookup file which expects the required symbols to be > exported using the compiler flags, that are mentioned in the Lib.gmk file. > The math library in AIX specifically, is a static archive. Doing a -lm wont > suffice, because when the symbols are looked up using dlsym or accessing > native code through Java, it will lead to failures. Hence we had to come up > with a list of symbols to allow math library symbols to be accesible. Also, > there are parts of libc library that are static too, and hence those symbols > also are present in this list. Without this change, the StdLibTest and > multiple other tests which make native function calls using FFI, fail with > NoSuchElementException. > For some context: `java.lang.foreign.Linker.nativeLinker().defaultLookup()` > returns a `SymbolLookup` that can be used to find symbols from the standard > library. > > We use a shim library that re-exports symbols from the standard library in > order to make them accessible through `dlsym`, since on e.g. Linux the > standard library .so files might actually be linker scripts which we don't > want to mess with. On Windows we use ucrtbase.dll + an auxiliary library that > contains the expansions of several 'inline' functions such as `printf`. I'm > not surprised to see that another platform like AIX requires more to make > this work as well. > > That said, I think we should limit the exported symbols to only those found > in the standard C library header files: https://en.cppreference.com/w/c/header Could the list of symbols be dynamically extracted at build time in some way or is the set stable enough to basically never change? ------------- PR Comment: https://git.openjdk.org/jdk/pull/16414#issuecomment-1804283653
