Skip to content

Port three small fixes from the DarthMDev and KartPad forks - #10

Merged
mitch030504 merged 5 commits into
openxr-workfrom
claude/project-thread-4jegmq
Oct 5, 2026
Merged

mitch030504 merged 5 commits into
openxr-workfrom
claude/project-thread-4jegmq

Conversation

@mitch030504

Copy link
Copy Markdown
Owner

Requested by Mitch · project thread

Before: when a Retro Rewind TLS write timed out after 15 s, the runtime closed the game's socket and cleared its SSL slot while the game was still using both, so a later connection could take the same slot under the old id. Disc reads were copied into guest RAM one byte at a time, every read reopened its file, and SZS archives were decompressed byte by byte through guest-memory accessors. The ImGui passes always recorded Vulkan debug labels, which an Adreno 740 Android driver crashes on.

After: a failed TLS write keeps the session but marks it failed, and later calls on it fail cleanly until the game closes it. Disc reads use memcpy into ordinary RAM, keep up to 8 file handles per thread, and SZS decoding uses direct pointers when the range is plain RAM. ImGui debug labels only exist when AURORA_GFX_DEBUG_GROUPS is defined, like the rest of Aurora's.

These come from the fork review (/mnt/project-files/fork-review/group-1.md in the project). Expect shorter loads, not more fps; the label change mostly matters for the Quest 3 build.

How: one commit per fix, each credited to its source.

Checked locally: network_ssl.cpp, dvd.cpp and egg_decomp.cpp pass clang++ -std=c++20 -fsyntax-only -Wall for Linux, with and without MKW_HEADSET_STEAM_FRAME, against mbed TLS 3.6.7 headers. A small harness for the cached DvdReadContract::ReadExact (repeat reads over more than 8 files, bad offsets, short reads, a file rewritten with a new mtime, a missing file, a second thread) passes, including under ASan and UBSan. Not built for the headset or played yet.

🤖 Generated with Claude Code

https://claude.ai/code/session_019rLZ24cFcqTNBmb4w2dpSq


Generated by Claude Code

DarthMDev and others added 5 commits October 5, 2026 15:54
When an mbed TLS write timed out, SslWrite called DeleteWiiSocket, which
closed the game's socket and cleared its SSL slot while the game still
held both ids. A later SSL_NEW could then hand the same slot out under
the old id. Now a failed write marks the session failed and shuts the
host socket down; later reads, writes and handshakes on it fail until the
game tears the session down itself.

Ported from DarthMDev/Wiicompiled a2ecc0f (network_ssl.cpp part) and
fcf8646, both on upstream PR patchzyy#258.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019rLZ24cFcqTNBmb4w2dpSq
Ported from chrissotraidis/wiicompiled (KartPad) ec14a0d.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019rLZ24cFcqTNBmb4w2dpSq
Ported from chrissotraidis/wiicompiled (KartPad) 97cdc9a.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019rLZ24cFcqTNBmb4w2dpSq
Ported from chrissotraidis/wiicompiled (KartPad) affabbd.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019rLZ24cFcqTNBmb4w2dpSq
Aurora's own debug groups only exist when AURORA_GFX_DEBUG_GROUPS is
defined, but the three ImGui passes (desktop overlay, host-frame overlay
and the headset settings panel) pushed and popped theirs in every build.
An Adreno 740 Android driver (512.676) faults in
vkCmdEndDebugUtilsLabelEXT on the first overlay pass (KartPad patchzyy#321); the
Quest 3 uses the same GPU. Gate all three like the rest of Aurora.

Ported from chrissotraidis/wiicompiled (KartPad) 420b828, extended to the
two extra passes this fork has.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019rLZ24cFcqTNBmb4w2dpSq
@mitch030504 mitch030504 self-assigned this Oct 5, 2026
@mitch030504
mitch030504 marked this pull request as ready for review October 5, 2026 16:13
@mitch030504
mitch030504 merged commit 58b6d81 into openxr-work Oct 5, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants