On Fri, 14 Nov 2025 11:25:05 GMT, Matthias Baesken <[email protected]> wrote:
> The dead_strip linker option on macOS removes functions and data that are
> unreachable by the entry point or exported symbols.
> Setting it can reduce the size of some binaries we generate quite a lot, for
> example (product build, Xcode 15 is used) :
> (before -> after setting the option)
>
> 1.4M -> 1.1M images/jdk/lib/libfontmanager.dylib
> 264K -> 248K images/jdk/lib/libjavajpeg.dylib
> 152K -> 132K images/jdk/lib/libjli.dylib
> 388K -> 296K images/jdk/lib/liblcms.dylib
> 164K -> 128K images/jdk/lib/libzip.dylib
>
>
> and libjvm :
>
> 20M -> 18M images/jdk/lib/server/libjvm.dylib
> 146M -> 137M images/jdk/lib/server/libjvm.dylib.dSYM
SA is throwing an exception on the following:
` Klass classLoaderKlass = vm.getSystemDictionary().getClassLoaderKlass();`
This looks to be the first thing it is doing that involves looking up a hotspot
symbol. Possibly the issue is with symbols in general being stripped, or maybe
just this one was. The fact that only 3 tests failed makes me think it is some
very selective dead stripping of symbols that is the problem. Most SA
functionality does not use SystemDictionary().getClassLoaderKlass(), so if just
this one symbol was missing, that would explain why for the most part SA is
working.
On the assumption that it is just this one symbol, Hotspot declares an array of
InstanceKlass* for some critical classes:
InstanceKlass* vmClasses::_klasses[static_cast<int>(vmClassID::LIMIT)]
= { nullptr /*, nullptr...*/
};
In SA, the failed SystemDictionary.initialize() part of the stacktrace shows an
attempt to access this array for the ClassLoader InstanceKlass*
` classLoaderKlassField =
type.getAddressField(VM_CLASS_AT("ClassLoader_klass"));`
private static String VM_CLASS_AT(String name) {
return "_klasses[static_cast<int>(" + VM_CLASS_ID(name) + ")]";
}
And in vmStructs we have:
` static_field(vmClasses, VM_CLASS_AT(ClassLoader_klass),
InstanceKlass*) `
I think this vmStructs static field is getting dead stripped.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/28319#issuecomment-3549555596
