Restrict special handling of MemorySegment methods to only the getters and setters #20696
Conversation
Other MemorySegment methods do not require special handling. Signed-off-by: Nazim Bhuiyan <[email protected]>
Peeking of MemorySegment method callers are necessary for obtaining the known object info of the layout argument that is used for inlining the VarHandle operation methods. This fix restricts the peeking only to those methods that call the MemorySegment getter and setter methods, and reduces unnecessary peeking. Signed-off-by: Nazim Bhuiyan <[email protected]>
|
@jdmpapin I'd appreciate a review from you as this mainly affects Inliner behaviour. |
|
Jenkins test sanity all jdk21 |
|
Jenkins test sanity win jdk21 |
|
AIX sanity.functional failed "OSCache Test" in shrtest_aix_1. Port layer error code -174 "invalid argument" has been observed in #16448. There was also an assertion failure later in the test, and a similar assertion failure has been observed in #10275. Based on the test output, I think it's related to the destruction of semaphores in response to shared memory errors. There were a few other error codes from the port library besides -174, but it's all related to shared memory and the port library, and potentially the machine setup. I'm confident those errors are unrelated to this change x86-64 Mac sanity.functional didn't run because it failed to fetch the merge commit x86-64 Mac sanity.openjdk didn't run because curl couldn't connect to artifactory z Linux sanity.openjdk failed DeflateIn_InflateOut in jdk_util_1. This doesn't seem to be a known issue. Can you look into this one please, @nbhuiyan? Jenkins test sanity xmac jdk21 |
|
There are no windows build machines. It looks like it's been that way for nearly two weeks |
This was also mentioned in #20376 (comment) last month. I don't think this particular change could cause this failure. I still launched a 20X Grinder run for |
|
Oh, it is a known issue after all: #14948 For some reason searching the issues for "DeflateIn_InflateOut" returned no results, but the issue shows up if I instead search for "DeflateIn InflateOut" with space instead of underscore |
|
So all test failures are known issues, and the windows build failure is due to a lack of windows build machines. Merging |
|
I have aborted the Grinder job so that we do not keep a zlinux machine occupied for a known issue. |
This PR does the following: