Repository navigation
Conversation
|
Note for the maintainers: #3499 is an earlier open PR implementing these same two exports — it carries @screamm's commit from #3342 unchanged, is CLEAN with all checks green, and has dual-platform verification (the author's Windows run in #3342 plus an independent Linux full-suite pass, 829/829 including the Vulkan suites). This PR looks like a clean independent implementation of the same issue, so whichever you prefer is fine — flagging only so the duplication is visible and one review isn't wasted. |
|
Closing as a duplicate: #3499 (opened earlier, mergeable, CI green) already implements Thanks for the work and the test. If #3499 turns out not to be the one that lands, or you find a gap in it, feel free to reopen this or comment there. |
What
_Cnd_signaland_Cnd_init_with_namenow operate on the libc-owned condition objects used by_Cnd_wait,_Cnd_broadcastand_Cnd_init._Cnd_signalwakes one waiter;_Cnd_init_with_nameinitializes the same object as_Cnd_initand ignores the optional name. TechnicalDebt records that the inferred ABI and behavior still need console confirmation.Commit
f6266acb8carries David's implementation from blawesom's #3499, retaining David as commit author and the cherry-pick source. The follow-up test commits preserve the two-waiternotify_oneregression coverage, bound each waiter-state wait to five seconds, and set the CTest timeout to 20 seconds so a broken signal path cannot leave this test hanging indefinitely.Tested
guest_ampr_threads_tests;guest_ampr_threadspassed 3/3 consecutive runs on this head.python3 tools/check_conventions.py --base upstream/mainandgit diff --check upstream/main...HEADpassed.Checklist
main; the overlapping implementation from feat(libSceAmpr): implement _Cnd_signal and _Cnd_init_with_name #3499 is carried and credited in this PRNotImplemented_nid_no_patch); silent stubs are listed in TechnicalDebtFixes #3342