Our sandbox passed all 50 Python tests for jiayx/d2r-mobile-build, while installation still requires matching game resources and Apple signing. Use this template if you already have those resources and want private GitHub Actions builds. Skip it if you expect an IPA to include everything needed to play. The README says the former community resource download is unavailable. Project installation instructions.
What people use d2r-mobile-build for
jiayx/d2r-mobile-build builds iOS and macOS apps through GitHub Actions and publishes them in your private repository. Its controls let you choose an upstream branch, tag or commit, or build a source fork. That makes the template relevant to existing users maintaining their own packages and developers testing a particular revision. The following issue reports show how people used the resulting packages. Build configuration.
| Who used it | What they did | Recorded outcome |
|---|---|---|
| GitHub user xinerqu | Ran a released IPA in LiveContainer | Reported successful gameplay and no first-launch shader compilation; the maintainer attributed this to packaging precompiled shaders. LiveContainer report in issue #11 |
| GitHub user zsc8917zsc | Enabled increased-memory-limit while signing | Reported stable operation and peak memory around 3.2 GB. The maintainer subsequently added the entitlement request. Signing discussion in issue #18 |
| GitHub user goldeye0351 | Tried experimental Apple TV deployment using two Macs | Reported progress above 70% and memory usage exceeding 100 GB. Deployment report in issue #21 |
The Mac package also addresses a specific request. In issue #8, a user said the phone screen was small and an external monitor separated the picture from touch input. Another participant wanted a Mac version too. The maintainer added macOS packages and recommended building from the template. GitHub recorded 449 stars and a November 15, 2026 last push in the supplied repository snapshot. Mac request, repository facts.
What happened when we ran it
Our run at commit 15861f9 passed 50 pytest tests in 7 seconds on November 15, 2026. Installation succeeded in 10 seconds, and the recorded build step succeeded in 1 second. These measurements describe the repository's Python tooling. For someone editing its build scripts, they provide a concrete result tied to a checkout rather than a general impression of the project's activity.
| Our recorded check | Result |
|---|---|
| Install | Succeeded in 10 seconds |
| Dependencies | 40 recorded, 40 MB on disk |
| Build | Succeeded in 1 second |
| Tests | Succeeded in 7 seconds, 50 passed and 0 failed of 50 |
| pip-audit | 0 known vulnerabilities |
| Checkout | 62 files, approximately 4,499 source lines, 0.3 MB |
| Repository signals | 12 CI workflow files, a tests directory, no Dockerfile |
The sandbox configuration recorded 3 CPUs, 8 GB RAM and image label ghcr.io/astral-sh/uv:python3.12-bookworm. A separate hands-on attempt failed its safety step with the exact error: "unsafe repo: none of compose.yaml, compose.yml, docker-compose.yaml, docker-compose.yml exists (rule: a compose file is required)". The documented project route uses GitHub Actions.
The hands-on header describes a Docker Compose deployment on November 15, 2026, with every port bound to 127.0.0.1. The failed safety step is the recorded outcome of that attempt; the teardown record says "teardown: ok". Reading those records together prevents the header from being mistaken for a successful application deployment.
Where people get stuck: resources, signatures and the wrong package
The former Google Drive resource download is unavailable, according to the README and issue #36. The maintainer says the account sharing the files was banned. Commenters ask for complete Data and Legacy packs, including one request that explicitly distinguishes game resources from an IPA. Existing matching resources remain reusable, so resource ownership is the first practical decision before starting a build. Resource status in issue #36.
| Symptom | What the record supports | Action supported by that record |
|---|---|---|
| Feather reports 0xe800801c | Issue #11 traces the reported failure to zsign handling of Guest.framework | Use a package containing the documented Guest ad-hoc signature repair, then re-sign it. Feather repair |
| Mac shows purple graphics with sound | Issue #14 identifies omitted shader reflection JSON | The historical repair shipped in v1.0.17 macOS build 8.1; ppstars reported normal graphics after replacing the app and its supplied metadata. Mac repair |
| LiveContainer cannot find Data | Issue #11 describes resources placed outside the guest app's container | Put the resources in the guest app's data/Documents location. The user then reported success. Resource-container exchange |
| M1 iPad exits when the naming keyboard opens | Issue #16 records a LiveContainer-specific reproduction | wwoodfox reported normal gameplay after switching to D2RMobile-v1.0.17-build.9.1-ios18.ipa. Keyboard workaround |
The Mac repair has a useful before-and-after record. ppstars reported a MacBook Air with Apple M4, 16 GB RAM and macOS 27.0.1, checked package and resource hashes, and described audio and click feedback behind purple graphics. After replacing the app and copying the supplied metadata, the user reported normal graphics. The replacement contained 7,212 .metallib files and 6,928 .json files; the earlier build 1.1 contained zero .json files. Package comparison.
The Feather report also identifies a specific packaging problem. Ti800 reproduced zsign's failure to insert Guest's signature load command and reported successful iPhone installation after pre-signing Guest. The maintainer then added the ad-hoc signature and retained its signature resources during packaging. xinerqu's later LiveContainer success is another user's report in the same thread, with a different installation context. Signing reproduction and maintainer response.
Updates preserve resources only when they still match
The README instructs iOS users to back up saves, keep the app identifier unchanged and install over the existing app. Previously imported matching Data and Legacy resources remain reusable. On macOS, users should quit the app, back up saves and replace the same app variant while keeping the resource folder in its established location. Update instructions.
The source controls are useful when maintaining private builds. source_ref accepts a branch, tag or commit, and source_repository selects a fork. A private source repository requires a SOURCE_TOKEN with Contents read access. Our recommendation is to record those settings alongside the package name when reporting a problem, so the report identifies the source and platform variant involved. Repeated runs receive separate release and file names. Workflow controls.
The current Mac instructions require the resource folder to contain build_manifest.psv and all 18 .data files inside data. If the manifest is missing, the README directs users to extract the complete matching archive again. Issue #14's metadata-copy procedure belongs to its historical repair; a fresh installation should follow the current folder instructions. Mac resource structure.
Signing requires preserving the memory entitlement
The IPA requests com.apple.developer.kernel.increased-memory-limit, and the README tells the signing tool to preserve it in the final signature and provisioning profile. The capability also needs enabling for the App ID. This makes the final signing configuration part of installation, even when the downloaded package already carries the request. The README lists Sideloadly, iloader, AltStore Classic and SideStore as installation options. Signing instructions.
In issue #18, zsc8917zsc reported crashes during portals, area changes and sometimes saving, as well as busy scenes. The reporter later said enabling increased-memory-limit helped. The maintainer added the request to both v2.2.0 iOS IPAs and completed packaging checks. The same response explains that removing the entitlement during re-signing makes the bundled request ineffective. Crash and entitlement discussion.
The packages include checksum and build-detail files: SHA256SUMS.txt and build-report.json for iOS, with corresponding Mac files. Those give users package identifiers to include in a report. Keep the package name, signing route and source revision together when comparing results with an issue comment, particularly when the comment describes a repair in an older release. Release-file descriptions.
English instructions exist; detailed repairs also appear in Chinese
The repository provides English and Simplified Chinese READMEs, and several detailed troubleshooting exchanges are in Chinese. The Mac purple-screen investigation and memory-entitlement discussion contain the package names, symptoms and follow-up results needed to understand those repairs. English-speaking readers can use the English setup instructions and translate the relevant issue comments when matching a specific failure. Project instructions, Mac investigation.
Issue #16 also contains an English reproduction report followed by troubleshooting of the M1 iPad keyboard crash. Across these exchanges, the useful reporting pattern is consistent: name the device, operating system, package, signing route and failing step. The Apple TV discussion illustrates the consequence of leaving out the step: the maintainer asks which tool owns the progress indicator before diagnosing the reported memory usage. Keyboard discussion, Apple TV follow-up.
d2r-mobile-build vs upstream Xcode and NO_XCODE builds
The jiayx template uses GitHub Actions, while upstream documents a full local Xcode build and experimental shell packaging. Choose between them according to where you want preparation and compilation to happen. Signing tools handle the later installation stage. Keeping these jobs separate helps explain why replacing a signing tool and replacing a build route answer different problems. jiayx workflow, upstream route table.
| Route or tool | Reason to choose it | Consequential difference |
|---|---|---|
| jiayx GitHub Actions template | You want private iOS or macOS source builds | Uses platform prepared input and includes precompiled Metal shaders. Template instructions |
| Upstream D2R Mobile with Xcode | You want upstream's full local Mac route | Upstream's supplied route table calls this its main, fully verified route. Upstream route table |
| Upstream NO_XCODE packaging | You want to prepare content from your own game files | When the computer lacks Apple's shader compiler, the guide says the phone compiles shaders on first launch. NO_XCODE guide |
| Sideloadly, AltStore or SideStore | You need an installation option distinct from Feather/zsign | Issue #11 says these tools do not use zsign and were unaffected by that reported failure. Signing comparison |
The similarly named d2r-mobile/d2r-mobile-build describes Ubuntu GitHub Actions builds using the NO_XCODE approach. This review concerns jiayx/d2r-mobile-build, whose README describes private iOS and macOS source builds. Read the owner name alongside the repository name before borrowing a configuration or interpreting a release-file description. The two supplied pages describe different workflows. Other build repository, jiayx instructions.
What the other reviews get wrong
The October Quriosity-agent audit's PREPARED_INPUT Secret instructions describe an older setup, while the supplied README uses the prepared_input workflow field. Its first-launch shader discussion also predates the packaging change described in issue #11. The current README says the IPA includes precompiled Metal shaders, and xinerqu reported starting without that compilation step. October audit, shader follow-up.
The supplied upstream skill names D2RMobile-shell.ipa, but the NO_XCODE guide names shell-build-input ZIP files and calls the shell build input rather than an installable app. Follow the guide's distinction when choosing an input. The October audit's community-download references also need the later resource-status update: issue #36 and the README report that the former Google Drive download is unavailable. Skill snapshot, shell guide, resource discussion.
OpenTrend describes 3.8 W during active combat, whereas upstream's measurement conditions say the phone was standing in the Act I camp at 50% brightness. Upstream also derives battery life from power consumption and battery capacity rather than a full discharge session. Its dated iPhone 17 Pro Max measurement should retain those conditions. OpenTrend's flawless-performance language also sits poorly beside the specific crash and graphics reports described above. OpenTrend description, upstream measurement conditions.
Questions people ask
The README says the former community Data and Legacy download is unavailable, and existing matching resources can be reused. That answers the immediate resource question raised in issue #36: an updated IPA and complete game resources are separate requirements. The following questions come from the supplied issue discussions and research synthesis, with answers tied to the relevant records. Resource instructions.
Where can I get the complete Data and Legacy resource packs now that the old download links are unavailable?
The README directs resource-availability discussion to issue #36 and tells existing owners to reuse matching files. Commenters there ask for complete Data and Legacy packs rather than another IPA. Upstream's NO_XCODE guide describes preparing content on your computer from your own game files, which is a different route from downloading a ready-made community resource archive. Resource requests, own-files preparation.
How do I run the game in LiveContainer and put its resources in the correct folder?
Issue #11 directs resources into the guest app's data container and Documents location. wuziwei had created a separate folder under On My iPhone and reported that the app could not load the resources. xinerqu explained the container location, and wuziwei replied that the problem was solved. That exchange supplies the container-specific correction to the ordinary installation instructions. LiveContainer resource exchange.
Why does Feather signing fail with installation error 0xe800801c?
Issue #11 attributes the reported failure to zsign's handling of Guest.framework's signature load command. Ti800 demonstrated successful installation after pre-signing Guest, and the maintainer incorporated an ad-hoc signature and retained its signature resources during packaging. The issue identifies the affected original IPA as D2RMobile-v1.0.16-build.5.1.ipa and the subsequent repair release as v1.0.17-build.8.1. Feather investigation.
Why does my iPhone 14 Pro Max sometimes crash during gameplay?
Issue #18 records crashes during portals, area changes, busy scenes and sometimes saving or exiting. The reporter suspected memory pressure and later reported improvement after enabling increased-memory-limit while signing. The maintainer added the entitlement request to v2.2.0 packages and asked for the signing method and fresh diagnostics if the crashes continued. iPhone crash discussion.
Does increased-memory-limit signing help prevent crashes during portals, area changes and saving?
zsc8917zsc reported stable operation after enabling the permission. The maintainer's follow-up instructs users to preserve it in the final signature and provisioning profile, because a re-signing tool can remove the request. For installation, the actionable step is to retain that capability rather than assume downloading a package with the request completes the signing configuration. Entitlement follow-up.
Why does opening the character-name keyboard crash the game on an M1 iPad in LiveContainer?
Issue #16 records normal gameplay after wwoodfox switched to D2RMobile-v1.0.17-build.9.1-ios18.ipa. The maintainer explained that this variant uses a different Metal path and that iOS 18 names its minimum supported system. The README now suggests trying the -ios18.ipa variant for this M1 LiveContainer symptom, including on iPadOS 26 or later. Keyboard workaround, README recommendation.
Can an Apple TV version be provided?
The README lists an experimental v2.2.0 Apple TV package targeting third-generation Apple TV 4K with A15, tvOS 26 or later and an extended gamepad. Its workflow uses a standalone prepared input of about 226 MB. The project describes native compilation and package checks as completed steps and links a dedicated TV guide for requirements, building and installation. Experimental TV instructions.
What causes the Apple TV setup to exceed 100 GB of memory usage around 70% progress?
Issue #21 records the symptom after attempts on two Macs, including an attempt after freeing Apple TV storage. The maintainer asks whether the progress belongs to signing, installation, resource import or game startup, and requests the process name and logs. Those details distinguish a Mac-side preparation problem from a game-runtime memory problem. Apple TV troubleshooting request.
Will this repository support other games or collect similar native-port projects?
The maintainer says this repository focuses on d2r-mobile workflows and has no plan to expand to other games or collect similar projects. That response answers xinerqu's question in issue #11. Choose the template for this upstream project's build workflow; the supplied response gives it a specific scope rather than presenting it as a catalogue of native game ports. Maintainer scope statement.
Is D2R Mobile just Ryujinx in a wrapper?
LNPANDY's X post distinguishes offline conversion of game code and shaders into native Apple Mach-O binaries and .metallib files from hardware emulation at runtime. That is the supported, attributed explanation in the supplied evidence. It addresses the execution-model question raised by the post, while the jiayx README describes the surrounding source-build and packaging workflow. Offline-conversion explanation, build-template description.
What does the open-source D2R Mobile port bring to iPhone?
The supplied Bilibili listing asks what the port brings to iPhone, while jiayx's README documents the concrete installation path for its packages: sign the IPA, import matching resources and open the app. It specifies A14 or M1 and newer hardware with at least 6 GB memory, with additional matching Legacy resources for A14 through A16 devices. Bilibili listing, device requirements.
Why does d2r-mobile run without an x86 emulator or JIT runtime?
LNPANDY's X post describes converting game code and shaders offline at build time into native Apple binaries and shader libraries. The post distinguishes that preparation from emulating hardware while the game runs. This attributed distinction answers the supplied question using the available passage, without adding an instruction-rewriter or host-services explanation. jiayx's documented role is to build and package upstream source. Execution-model explanation, template role.
What we have not tested yet
Our measurements cover Python tooling at commit 15861f9, not physical-device acceptance of an IPA or Mac app. The latest release fetched on November 15, 2026 is v2.3.2-build.21.1; its notes explicitly say the build has not been tested on a physical device. The earlier v2.3.0 maintainer statement belongs to that earlier release. Historical user successes retain their named packages and devices. Latest release notes, earlier release discussion.
The one-second build entry is the harness step, not an Apple compilation or installation duration. The image label and sandbox configuration do not establish actual runtime or operating-system versions without raw toolchain logs. The Compose attempt supplied no gameplay answers or screenshot records. pip-audit's zero-known-vulnerability result concerns the Python dependency scan; prepared inputs and final applications received no security audit here.
Signing compatibility across tools, versions and account types, resource import on hardware, long sessions, save recovery, thermals, frame rates and battery life remain outside our run. The README says signing support varies by tool and version, and increased-memory-limit guarantees neither a fixed allowance nor elimination of every crash. The supplied records contain no confirmed replacement complete resource download. Issue #21 leaves the Apple TV memory cause unresolved, and TV saves use purgeable Caches storage.
The supplied page-17 passage covers performance conditions rather than the instruction-rewriter and host-services details previously asserted in this FAQ. The X post supplies an attributed execution-model explanation, not a jiayx-versus-emulator benchmark. The YouTube samples concern character builds or other mods; the relevant Bilibili record is a listing, and the supplied evidence includes no Reddit gameplay report. Upstream measurement passage, X explanation, video sample, Bilibili listing.
A useful next device record would identify the exact release, hardware, OS, signing tool and retained entitlements, then demonstrate resource loading, character creation, a portal, an area change, saving and reopening. That is a proposed acceptance sequence, not a test performed for this review. Preserve working resources and saves before replacing the application, as the project's update instructions recommend. Update instructions.


