8240254: Build is broken when cds is disabled after JDK-8236604(Internet mail)
Calvin Cheung
calvin.cheung at oracle.com
Sat Feb 29 02:27:02 UTC 2020
Hi Jie,
Before seeing your email, I've filed
https://bugs.openjdk.java.net/browse/JDK-8240257 per Yumin's request
since he was having some problem logging into JBS.
If the issue has been fixed, can you close the above bug with your
latest message in this email thread.
thanks,
Calvin
On 2/28/20 6:18 PM, jiefu(傅杰) wrote:
> Yes. It passed on my computer.
>
> Apple clang version 11.0.0 (clang-1100.0.33.16)
> Target: x86_64-apple-darwin19.0.0
>
> Because SystemDictionary::load_shared_class_misc(InstanceKlass* ik, ClassLoaderData* loader_data, TRAPS) is guarded by #INCLUDE_CDS.
> So I think it OK.
>
>
>
> From: Yumin Qi <yumin.qi at oracle.com>
> Date: Saturday, February 29, 2020 at 10:03 AM
> To: "jiefu(傅杰)" <jiefu at tencent.com>, Claes Redestad <claes.redestad at oracle.com>, "hotspot-runtime-dev at openjdk.java.net" <hotspot-runtime-dev at openjdk.java.net>
> Subject: Re: 8240254: Build is broken when cds is disabled after JDK-8236604(Internet mail)
>
>
> OK, I will file bug for it. Did you get your build passed? How come the two versions not found by compiler?
>
> Thanks
>
> Yumin
> On 2/28/20 5:58 PM, jiefu(傅杰) wrote:
> Hi Yumin,
>
> It's too late.
> Sorry, I missed your comments.
>
> I think you can file a new bug to fix it.
>
> Thanks a lot.
> Best regards,
> Jie
>
> From: Yumin Qi <yumin.qi at oracle.com><mailto:yumin.qi at oracle.com>
> Date: Saturday, February 29, 2020 at 9:49 AM
> To: Claes Redestad <claes.redestad at oracle.com><mailto:claes.redestad at oracle.com>, "jiefu(傅杰)" <jiefu at tencent.com><mailto:jiefu at tencent.com>, "hotspot-runtime-dev at openjdk.java.net"<mailto:hotspot-runtime-dev at openjdk.java.net> <hotspot-runtime-dev at openjdk.java.net><mailto:hotspot-runtime-dev at openjdk.java.net>
> Subject: Re: RFR: 8240254: Build is broken when cds is disabled after JDK-8236604(Internet mail)
>
>
> HI, Claes and Jie
>
> Thanks for reporting this and gave a quick fix. Yes, this function should be guarded by INCLUDE_CDS:
>
> in systemDictionary.hpp:
>
> static void load_shared_class_misc(InstanceKlass* ik, ClassLoaderData* loader_data, TRAPS) NO_CDS_RETURN;
>
> in systemDictionary.cpp:
>
> iff -r 624d51be7acd src/hotspot/share/classfile/systemDictionary.cpp
> --- a/src/hotspot/share/classfile/systemDictionary.cpp Fri Feb 28 12:49:37 2020 -0800
> +++ b/src/hotspot/share/classfile/systemDictionary.cpp Fri Feb 28 17:46:22 2020 -0800
> @@ -1918,6 +1918,7 @@
> }
> #endif
>
> +#if INCLUDE_CDS
> void SystemDictionary::quick_resolve(InstanceKlass* klass, ClassLoaderData* loader_data, Handle domain, TRAPS) {
> assert(!Universe::is_fully_initialized(), "We can make short cuts only during VM initialization");
> assert(klass->is_shared(), "Must be shared class");
> @@ -1948,6 +1949,7 @@
> add_to_hierarchy(klass, CHECK);
> assert(klass->is_loaded(), "Must be in at least loaded state");
> }
> +#endif
>
> bool SystemDictionary::resolve_wk_klass(WKID id, TRAPS) {
> assert(id >= (int)FIRST_WKID && id < (int)WKID_LIMIT, "oob");
>
>
>
> The function needs guarded too or in Not CDS build, it will have two versions.
>
>
>
> Thanks
>
> Yumin
>
>
>
>
> On 2/28/20 5:24 PM, Claes Redestad wrote:
> Oops, wrong method:
>
> diff -r f227e770495f src/hotspot/share/classfile/systemDictionary.hpp
> --- a/src/hotspot/share/classfile/systemDictionary.hpp Fri Feb 28 15:30:29 2020 -0800
> +++ b/src/hotspot/share/classfile/systemDictionary.hpp Sat Feb 29 02:24:12 2020 +0100
> @@ -600,7 +600,7 @@
> const ClassFileStream *cfs,
> TRAPS);
> // Second part of load_shared_class
> - static void load_shared_class_misc(InstanceKlass* ik, ClassLoaderData* loader_data, TRAPS);
> + static void load_shared_class_misc(InstanceKlass* ik, ClassLoaderData* loader_data, TRAPS) NO_CDS_RETURN;
> static InstanceKlass* load_shared_boot_class(Symbol* class_name,
> TRAPS);
> static InstanceKlass* load_instance_class(Symbol* class_name, Handle class_loader, TRAPS);
>
>
> On 2020-02-29 02:21, Claes Redestad wrote:
>
>
> Hi,
>
> I think we typically add NOT_CDS_RETURN to the declaration of the method
> in systemDictionary.hpp:
>
> diff -r f227e770495f src/hotspot/share/classfile/systemDictionaryShared.hpp
> --- a/src/hotspot/share/classfile/systemDictionaryShared.hpp Fri Feb 28 15:30:29 2020 -0800
> +++ b/src/hotspot/share/classfile/systemDictionaryShared.hpp Sat Feb 29 02:20:09 2020 +0100
> @@ -271,7 +271,7 @@
> }
>
> static void update_shared_entry(InstanceKlass* klass, int id);
> - static void set_shared_class_misc_info(InstanceKlass* k, ClassFileStream* cfs);
> + static void set_shared_class_misc_info(InstanceKlass* k, ClassFileStream* cfs) NO_CDS_RETURN;
>
> static InstanceKlass* lookup_from_stream(Symbol* class_name,
> Handle class_loader,
>
> This should fix your issue with less clutter.
>
> Looks good and trivial either way to me.
>
> Thanks!
>
> /Claes
>
> On 2020-02-29 02:06, jiefu(傅杰) wrote:
>
>
> Hi all,
>
> JBS: https://bugs.openjdk.java.net/browse/JDK-8240254
>
> Build is broken when cds is disabled.
>
> It might be fixed by
> ------------------------------------
> diff -r f227e770495f src/hotspot/share/classfile/systemDictionary.cpp
> --- a/src/hotspot/share/classfile/systemDictionary.cpp Fri Feb 28 15:30:29 2020 -0800
> +++ b/src/hotspot/share/classfile/systemDictionary.cpp Sat Feb 29 08:42:41 2020 +0800
> @@ -1941,7 +1941,9 @@
> }
>
> klass->restore_unshareable_info(loader_data, domain, THREAD);
> +#if INCLUDE_CDS
> load_shared_class_misc(klass, loader_data, CHECK);
> +#endif
> Dictionary* dictionary = loader_data->dictionary();
> unsigned int hash = dictionary->compute_hash(klass->name());
> dictionary->add_klass(hash, klass->name(), klass);
> ------------------------------------
>
> Could you please review it and give me some advice?
>
> Thanks a lot.
> Best regards,
> Jie
More information about the hotspot-runtime-dev
mailing list