With APS5_ENABLE_TRACY=ON libSceAgcDriver.prx imports libTracyClient.dll.
The test copies the libraries it is given into its own libs/ directory,
so on Windows the relinked app fails to load guest_loadexec_host.prx
with error 126. Pass the Tracy client along with the other libraries
when it is built.
main's journal format is already 5, so this branch's bump to 5 no
longer marks the journals written before its polygon-mode field.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
PA_SU_SC_MODE_CNTL with POLY_MODE set to a point or line type was refused,
and every triangle was drawn filled. The state decode now takes the polygon
type of the faces that are not culled into State::polygonMode, ignores the
types when POLY_MODE is clear or the primitives are points or lines, and
keeps refusing front and back types that differ while both faces are drawn.
Points are drawn one pixel wide (PA_SU_POINT_SIZE 0x00080008 inside
PA_SU_POINT_MINMAX), the size VK_KHR_maintenance5 gives a point whose
shader does not write PointSize, and lines one pixel wide (PA_SU_LINE_CNTL).
The device enables fillModeNonSolid and maintenance5 when they are
available. The polygon mode joins the pipeline, library and draw keys and
the pipeline journal (format 5).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A store to the trimmed tail of a written buffer would be dropped instead
of failing as an alias of the render target, so only read-only buffers,
such as per-attribute vertex fetches, are trimmed.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
shadPS4 declares sceNpSetGamePresenceOnline(req_id, online_id, presence)
and returns INVALID_ARGUMENT for request id 0 or a null online id. The
export took (user id, timeout), so a caller's arguments landed in the
wrong registers.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
An R128 image access reads only words 0-3 of its T#; the recompiler
passes words 4-7 as zero, so the decoder saw MAX_MIP 0 and refused a
128-bit T# with BASE_LEVEL above 0 ("starts past the surface's last mip
level at levels that would move the surface's own"). Streamed textures
such as a 2048x2048 BC3 T# naming its resident levels 2..11 hit this.
Where a T# with words 4-7 zero would be refused that way, it is now read
as a 128-bit T#: MAX_MIP = LAST_LEVEL, with levels BASE..LAST described
as their own chain (size and MIN_LOD rebased), but only for 1D and 2D
T#s whose rebased levels have the same offsets, sizes and tail slots as
in the full chain. Everything else still throws as before.
AI-assisted: yes
Rewrite the technical debt entry to list what is unverified on the
PS5 and what still throws, with the source each behaviour follows,
and drop the descriptions of behaviour that is no longer a gap. The
entry is now under the 1000 character limit.
A thread blocked in sceZlibWaitForDone was never woken by
sceZlibFinalize, so it waited for its whole timeout or forever.
Finalization now wakes every waiter, and a waiter that sees the
service finalized throws, since no source says what the console
returns. shadPS4 leaves such a waiter blocked.
sceZlibWaitForDone threw for any timeout and when no finished request
was pending. Wait on a condition variable that sceZlibInflate signals:
a null timeout blocks until a request finishes, and a timeout returns
0x81120027 (timed out) on expiry, as in shadPS4. The timeout is read
as microseconds, as SharpProspero documents for the PS5; shadPS4
reads milliseconds.
sceZlibFinalize threw while a finished request's result was unread.
Discard the stored results and the finished IDs not yet returned by
sceZlibWaitForDone instead. shadPS4 also returns success here and
never reports the abandoned requests; its next initialization clears
them. An ID from before finalization reports 0x81120002 afterwards.
sceZlibGetResult threw on a second call for the same request and on
an ID it did not hold. Return the same length and status on every
call, and 0x81120002 (not found) for an unknown ID, as shadPS4 does.
Null output pointers still throw.
sceZlibInflate threw when an earlier request's result had not been
read. Accept the request instead: every finished request keeps its
result, and sceZlibWaitForDone returns the finished IDs in submission
order, as shadPS4's done queue does. The PS5 completion order is
unverified and recorded in the technical debt entry.
Keep finished requests in a map keyed by request ID and their IDs in
a completion queue instead of a single optional result. This prepares
for several outstanding requests.
sceZlibWaitForDone now pops the completion queue, so it returns each
finished request once and no longer depends on whether its result was
read, as in shadPS4's done queue. A second wait for the same request
throws as before when nothing else is pending.
Both CTest entries run the same executable, and std::rand() starts from
the same seed in each, so parallel runs shared and removed each other's
files. The names now carry the mode and a random_device value.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01RbLgECLQ2bpDM7oe2urGZK
Dean