On Tue, 26 May 2026 22:25:26 GMT, Benjamin Peterson <[email protected]> wrote:

>> Benjamin Peterson has updated the pull request incrementally with one 
>> additional commit since the last revision:
>> 
>>   add comment before continue
>
> Thank you. Any suggestions for who to tag for a second review?

@benjaminp I am happy to review this as is. I'll just make a few observations 
which any help with documenting the output that appears in the listing.

This problem is specific to StubGenerator code blobs which are the only ones 
that include multiple stubs in a blob. Most of the StubGenerator blobs embed 
*managed* stubs whose details, including the blob and stub names, are declared 
in shared file `runtime/stubDeclarations.hpp`. A few (GC blobs) still include 
*unmanaged* stubs. Details of the latter, inlcuding names, are hard-coded 
somewhere in the generator code.

All StubGenerator stubs have an associated StubCodeDesc created with an 
associated stub group and stub name. Unmanaged stubs are not guaranteed to 
define these consistently. However, with managed stubs there is a clear way of 
identifying these values.

The group name is always `"StubRoutines"`. Arguably it ought to be something 
more useful, indicating that this is a StubGen stub and perhaps also 
identifying the actual group of StubGen stubs in the blob (initial, compiler 
etc) but we used `"StubRoutines"` because that was what was used in older 
releases and we were trying not to break tests or user code that processed 
diagnostic and log output.

The stub name is always uniformly derived from the `do_stub` declaration in  
`runtime/stubDeclarations.hpp`. So, for example for the preuniverse `fence` 
stub the relevant declaration line specifies `do_stub(preuniverse, fence)` and 
the stub name that results is `"fence_stub (stub gen)"`. Again, one might argue 
that this is not a useful format but i) it is uniform with the format used for 
non-StubGen blobs/stubs e.g. `"moitorenter_blob (C1 runtime)"` and ii) it only 
mildly perturbs the formats used in older releases.

Note that in the case of the stub you are testing for the declaration employs 
the previously chosen name `call_stub` so the resulting full name is 
`"call_stub_stub (stub gen)"`!

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

PR Comment: https://git.openjdk.org/jdk/pull/31007#issuecomment-4554084255

Reply via email to