8357119: Make Zero the default and only variant of HotSpot for iOS #35

TheShermanTanker · 2025-05-16T08:34:03Z

Currently, the Zero variant of HotSpot must be manually selected for mobile to compile properly for iOS, since iOS does not allow writeable and executable sections. Since Zero is, to my knowledge, the only valid HotSpot variant for iOS, this default should be handled by the build rather than require the developer to have specific knowledge about iOS and the mobile fork. I also took the opportunity to explicitly disallow selecting other JVM variants at all for iOS, this can be rolled back if my assumption turns out to be incorrect. While here, also touch up the build docs a little.

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8357119: Make Zero the default and only variant of HotSpot for iOS (Enhancement - P4)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/mobile.git pull/35/head:pull/35
$ git checkout pull/35

Update a local copy of the PR:
$ git checkout pull/35
$ git pull https://git.openjdk.org/mobile.git pull/35/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 35

View PR using the GUI difftool:
$ git pr show -t 35

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/mobile/pull/35.diff

Using Webrev

Link to Webrev Comment

bridgekeeper · 2025-05-16T08:34:43Z

👋 Welcome back jwaters! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2025-05-16T08:35:57Z

❗ This change is not yet ready to be integrated.
See the Progress checklist in the description for automated requirements.

mlbridge · 2025-05-16T08:39:23Z

Webrevs

johanvos · 2025-05-16T12:35:09Z

make/autoconf/hotspot.m4

@@ -74,6 +78,10 @@ AC_DEFUN_ONCE([HOTSPOT_SETUP_JVM_VARIANTS],
  JVM_VARIANTS=`$ECHO $JVM_VARIANTS_OPT | $SED -e 's/,/ /g' -e 's/minimal1/minimal/'`
  AC_MSG_RESULT([$JVM_VARIANTS])

+  if test "x$OPENJDK_TARGET_OS" = xios; then
+    VALID_JVM_VARIANTS="zero"


This should probably be guarded with a check on Device versus simulator. When running on an iOS simulator, it is technically possible to dynamically generate AND execute code, hence this is not restricted to zero.

Ah, I was not aware that one could run server on the iOS simulator. I can remove this part since my assumption seems to be incorrect.

johanvos · 2025-05-16T12:47:41Z

README.md

@@ -19,8 +19,8 @@ tracking.

 ### Pre-requisites
 Following are the prerequisites to build JDK on Mac targeting iOS:
-1. Download and unzip pre-built JDK24 to be used as boot JDK
-2. Download the [support zip](https://download2.gluonhq.com/mobile/mobile-support-20250106.zip) which contains an ios build for libffi and cups. Unzip it in the root directory of the project.
+1. Download and install JDK24 for macOS. You can use the JDK you just built targeting macOS in the above note instead, but this is not recommended.


out of curiosity, why is that not recommended?

Yes, I can't see why that would be a problem.

I was conservative and thought we wouldn't want to encourage using a bleeding edge JDK compiled off master as the boot JDK. If this is ok, I can remove that "Not recommended" note.

johanvos · 2025-05-16T13:01:27Z

README.md

 --with-libffi-include=<support-dir-absolute-path>/libffi/include \
 --with-libffi-lib=<support-dir-absolute-path>/libffi/libs \
 --with-cups-include=<support-dir-absolute-path>/cups-2.3.6
 --with-sysroot=/Applications/Xcode.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk \
 ```

+If you need to tell configure the path of the boot JDK, or if configure fails with an error saying
+it can't find the boot JDK, for instance if you downloaded a JDK in compressed archive form rather


minor: change "the boot JDK" to "a boot JDK"

I'll change this as soon as possible.

johanvos · 2025-05-16T13:03:13Z

@TheShermanTanker thank you for the PR. Looks good, some minor comments.
The biggest minor comment is probably about restricting the VM variant to Zero for all iOS builds. I believe there is value in building the server variant for the iOS simulator. While you can't distribute apps with it, it might help developers testing their apps quickly.

magicus · 2025-05-16T15:45:01Z

README.md

 --disable-warnings-as-errors \
 --openjdk-target=aarch64-macos-ios \
--with-boot-jdk=/Users/abhinay/.sdkman/candidates/java/24.ea.29-open \
--with-jvm-variants=zero \
 --with-libffi-include=<support-dir-absolute-path>/libffi/include \
 --with-libffi-lib=<support-dir-absolute-path>/libffi/libs \
 --with-cups-include=<support-dir-absolute-path>/cups-2.3.6


Configure arguments don't have to be absolute paths; configure will rewrite then to absolute path if they are relative.

magicus · 2025-05-16T15:47:40Z

@johanvos How can you test if you are building for the device or an emulator?

If this can/should be supported, then we also need to write something about it in the documents, not only update configure.

And if we can't detect the difference, then maybe most of this PR will need to be scrapped.

johanvos · 2025-05-16T17:05:50Z

@johanvos How can you test if you are building for the device or an emulator?

If this can/should be supported, then we also need to write something about it in the documents, not only update configure.

And if we can't detect the difference, then maybe most of this PR will need to be scrapped.

That is defined in the sysroot that is passed with configure. If you provide
--with-sysroot=/Applications/Xcode.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk you are building for the device.

If you provide
--with-sysroot=/Applications/Xcode.app/Contents/Developer/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/
you are building the the simulator.

If the latter SDK is selected, #ifdef TARGET_OS_SIMULATOR will return true when compiling code. (but that won't help for the makefiles to detect if we're building for device or simulator)

magicus · 2025-05-16T20:21:43Z

We can test for ifdefs by test compiling code, but it is not the most efficient way. Checking for if [[ SYSROOT =~ .*Simulator.* ]] seems simpler.

TheShermanTanker · 2025-05-17T04:07:40Z

@johanvos How can you test if you are building for the device or an emulator?

If this can/should be supported, then we also need to write something about it in the documents, not only update configure.

And if we can't detect the difference, then maybe most of this PR will need to be scrapped.

I still think defaulting to Zero for iOS is helpful, but I can remove the restricting of valid iOS JVM variants to only Zero since that is problematic.

johanvos · 2025-05-18T17:53:48Z

@johanvos How can you test if you are building for the device or an emulator?
If this can/should be supported, then we also need to write something about it in the documents, not only update configure.
And if we can't detect the difference, then maybe most of this PR will need to be scrapped.

I still think defaulting to Zero for iOS is helpful, but I can remove the restricting of valid iOS JVM variants to only Zero since that is problematic.

Defaulting to Zero for iOS sounds like a good idea indeed.

Make Zero the only valid JVM for hotspot.m4

40fb383

openjdk bot changed the title ~~8357119~~ 8357119: Make Zero the default and only variant of HotSpot for iOS May 16, 2025

openjdk bot added the rfr label May 16, 2025

Touch up README.md

0eb8eb0

johanvos reviewed May 16, 2025

View reviewed changes

magicus reviewed May 16, 2025

View reviewed changes

TheShermanTanker added 2 commits May 19, 2025 13:10

Improve README.md

ddd4e4f

Allow other JVMs again in hotspot.m4

3fed55c

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

8357119: Make Zero the default and only variant of HotSpot for iOS #35

8357119: Make Zero the default and only variant of HotSpot for iOS #35

TheShermanTanker commented May 16, 2025 •

edited by openjdk bot

Loading

bridgekeeper bot commented May 16, 2025

openjdk bot commented May 16, 2025

mlbridge bot commented May 16, 2025 •

edited

Loading

johanvos May 16, 2025

TheShermanTanker May 17, 2025

johanvos May 16, 2025

magicus May 16, 2025

TheShermanTanker May 17, 2025

johanvos May 16, 2025

TheShermanTanker May 17, 2025

johanvos commented May 16, 2025

magicus May 16, 2025

magicus commented May 16, 2025

johanvos commented May 16, 2025

magicus commented May 16, 2025

TheShermanTanker commented May 17, 2025

johanvos commented May 18, 2025

8357119: Make Zero the default and only variant of HotSpot for iOS #35

Are you sure you want to change the base?

8357119: Make Zero the default and only variant of HotSpot for iOS #35

Conversation

TheShermanTanker commented May 16, 2025 • edited by openjdk bot Loading

Progress

Issue

Reviewing

bridgekeeper bot commented May 16, 2025

openjdk bot commented May 16, 2025

mlbridge bot commented May 16, 2025 • edited Loading

Webrevs

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

johanvos commented May 16, 2025

Choose a reason for hiding this comment

magicus commented May 16, 2025

johanvos commented May 16, 2025

magicus commented May 16, 2025

TheShermanTanker commented May 17, 2025

johanvos commented May 18, 2025

TheShermanTanker commented May 16, 2025 •

edited by openjdk bot

Loading

mlbridge bot commented May 16, 2025 •

edited

Loading