Apply the import audits of ProsperoStore, vkQuake and Cloud Play

[AlxcBTW]

What

Closes #24, closes #25, closes #26. All three reports have the same cause. Each catalog title had a chain of one blocker. The import audits of its catalog build on upstream main ccf9ff90 show several absent imports. The miner reads loader lines and issue titles, not import audits, so this PR curates the chains in data/blockers.json. It has no rule change. These fixes share the curated blockers 3249 and pr-1841, so separate PRs would add the same ids twice.

• Blocker: #2908 ProsperoStore is fixed by boykopovar/AnyPS5#2945; its next stop is pr-1482 #24, ProsperoStore (PPSA99000). 2908 (sceKernelGetSystemSwVersion) gets status: resolved with #2945 as its fix PR. #2945 merged on 2026-10-11 at 07:49Z, but its body has no closing reference, so the issue is still open and mining linked #2945 only as a mention. The title's chain is now pr-1482 (first stop on main after #2945), then pr-1841.
• Blocker: #3251 shm_unlink does not block vkQuake (PPSA99010) #25, vkQuake (PPSA99010). #3251 (shm_unlink) goes to not_blockers. Its issue title names vkQuake, but vkQuake's catalog build does not import tPWsbOUGO8k. No other title in the snapshot is shown to stop there either. vkQuake's chain is now pr-1482 (its stop on main 9c5d67fb with #2967, which is merged), then 3249 (localeconv).
• Blocker: #3249 localeconv is one of nine absent imports of Cloud Play (PPSA99600), not its only stop #26, Cloud Play (PPSA99600). Its chain is now 3249, pr-1482, pr-1841. So it counts as frontier for 3249 but not as one fix away. 3249 becomes a curated blocker with area: libc. The mined area was libkernel, taken from the label of the open PR #1482, which the thread only mentions. Since a comment at 08:56Z named a seventh title, #3249 counted as a tracker, and on the current snapshot it had no titles at all (score 0). As a curated blocker it is ranked again.
• New pr-1841: getcontext, setcontext and makecontext. The only PR for them, #1841, was closed without merge. Its signature is 2VRU7xiqLO8 (setcontext), the line EVO Player (PPSA99039) stops at on main f60923bf on Windows. So EVO Player's mined chain picks it up as well.

Left out on purpose: Cloud Play's fnmatch, getservbyname and arc4random (no upstream issue or PR), and vkQuake's pthread_attr_* imports (#3165 is not a blocker on the map; it reads as a tracker). getprogname is no longer absent, because #3347 merged at 08:49Z, after the audits.

Source

• #2945: merged 2026-10-11T07:49:38Z, "Implement sceKernelGetSystemSwVersion for the startup path reported by ProsperoStore (PPSA99000)". closingIssuesReferences is empty, and #2908 is open.
• #2908 comment: "Linux (WSL2, GCC 13.3, llvmpipe), main ccf9ff90: ./app.elf: symbol lookup error: ./app.elf: undefined symbol: 04AjkP0jO9U", the same import on Windows main f60923bf, and the audit list with getcontext, setcontext and makecontext. Both commits contain #2945's merge commit 8ee2834a (compare shows them 96 and 63 commits ahead, 0 behind).
• #3251 comment: "does not import shm_unlink: … the NID tPWsbOUGO8k appears nowhere among the 335 NID#lib#module symbol strings". #2967 (merged 2026-10-10): "vkQuake (PPSA99010) | 04AjkP0jO9U (_umtx_op)" on main 9c5d67fb with the PR, on Linux and Windows.
• #3249 comment: "9 absent imports, localeconv is one of them". #3249 and #3310 name localeconv as Cloud Play's stop on main.
• #1841 (closed 2026-10-09, not merged) and its comments: setcontext on main 52ce9b8a, EVO Player on Windows main f60923bf, ProsperoStore and Cloud Play.

I also checked the claims independently. I downloaded the three catalog zips from the URLs in ps5-homebrew-catalog, and their sha256 values match the records. I then counted their NID#lib#module strings:

• vkQuake's eboot.bin has 335 of them. tPWsbOUGO8k is in neither the eboot nor libc.prx, while QuJYZ2KVGGQ (shm_open), 04AjkP0jO9U and 0hlfW1O4Aa4 are present.
• Cloud Play's eboot imports all nine NIDs of the audit.
• ProsperoStore's eboot imports 04AjkP0jO9U, dar03AOn+nM, 2VRU7xiqLO8 and GipcbdDM5cI.

In AnyPS5's progress.json, sceKernelGetSystemSwVersion and getprogname are done. localeconv, _umtx_op and the three context functions do not appear at all.

Local preview, from a live snapshot of 2026-10-11T09:01Z:

• pr-1482 moves from rank 3 to 2 (score 13 → 20). It gets ProsperoStore and vkQuake as frontier.
• pr-1841 is new at rank 9 (score 8).
• #3249 goes from score 0 to 4: frontier for Cloud Play, and listed for vkQuake.
• #2908 and #3251 leave the ranking.
• Open blockers go from 85 to 84, and ranked blockers stay at 77.
• Four titles change, in their open chains only: ProsperoStore, vkQuake, Cloud Play and EVO Player. No stages or claims change.

The CI preview uses the older snapshot of 07:55Z. That snapshot was taken before #3249 counted as a tracker and before EVO Player's setcontext comment. There:

• #3249 goes from score 12 to 10. Cloud Play stays frontier but is no longer one fix away, while SymphonyStation5 (PPSA17641) still is.
• pr-1482 moves from rank 4 to 2 (score 13 → 20).
• pr-1841 enters at rank 47 (score 2).
• #2908 and #3251 leave the ranking. Open blockers go from 79 to 78, and ranked blockers from 71 to 70.
• Three titles change: ProsperoStore, vkQuake and Cloud Play.

Checklist

• Every new stage, blocker or link has a public source; no invented title IDs, names or stages
• Stages count only for main, or main plus PRs that are merged now (no stages added)
• A new or changed mining rule comes with a fixture test in tests/ (no rule change)
• Tests, ruff check . and ruff format --check . pass locally
• I read the "Blocker map preview" in the CI job summary and it shows what I expected
• AI-assisted: yes

🤖 Generated with Claude Code