RFR: 8364360: Defining hidden class with no room in constant pool crashes the VM [v2]

[Coleen Phillimore]

On Mon, 27 Oct 2025 04:17:27 GMT, David Holmes <dholmes at openjdk.org> wrote:

>> Coleen Phillimore has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Test enhancement and comment.
>
> src/hotspot/share/classfile/classFileParser.cpp line 5528:
> 
>> 5526:     cp_size++;
>> 5527:     // Check for overflow.  cp_size is a u2.
>> 5528:     precond(sizeof(cp_size) == sizeof(u2));
> 
> Why do you need to assert this given `u2 cp_size = ...` is the declaration?

In case somebody changes it to int.  There used to be talk about doing this so then the overflow check might have to be different.

> test/hotspot/jtreg/runtime/ClassFile/HiddenClassesTest.java line 44:
> 
>> 42:         var cw = new ClassWriter(0);
>> 43:         cw.visit(V17, ACC_PUBLIC, "Hidden", null, "java/lang/Object", null);
>> 44:         for (int i = 0; i < 65530; i++) {
> 
> Why 65530? An empty class definition already has 12 CP entries when compiled by javac.

This is a magic number.  65536-5 gets CFE: class too large, 65536-7 doesn't get an CFE.  Only 65536-6 caused the overflow.  This is asm so asm may only be adding 6 entries.  I kept the test as ASM rather than using ClassFile API because it might be good to backport this.

> test/hotspot/jtreg/runtime/ClassFile/HiddenClassesTest.java line 50:
> 
>> 48:             MethodHandles.lookup().defineHiddenClass(cw.toByteArray(), false);
>> 49:             throw new RuntimeException("Test Failed: ClassFormatError expected.");
>> 50:         } catch (ClassFormatError cfe) {
> 
> It would be prudent to check that you get the expected CFE.

okay.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/27964#discussion_r2465450393
PR Review Comment: https://git.openjdk.org/jdk/pull/27964#discussion_r2465501998
PR Review Comment: https://git.openjdk.org/jdk/pull/27964#discussion_r2465481673