Install: Build v3.0.0-rc.4 fails compiling code_051.c with malformed _src/$_dst declaration #985

Open
opened 2026-11-16 00:42:58 +00:00 by BeanIssues · 5 comments
Collaborator

Reported by BobbyTortilla on Discord

What you used

Longshot Launcher v1.0.0-rc.9 (portable), building with the Longshot builder.

Version and channel

Build v3.0.0-rc.4 (release candidate). The same failure happened on launcher rc.7 and rc.9. It also reproduces with Longshot nightly-287-0407ae24.

Where it stopped

Build stopped during Compile at 53% (make_game.py / build.py, exit status 1), while compiling code_051.c on the rc.4 attempt. On nightly-287 it stopped compiling code_026.c after 728 seconds total; that file took 7 seconds compiling alone.

Error

On rc.4:

E:\OOT Remake\LongshotLauncher-data\work\sa\build\obj\code_051.split.c:253203:51: error: use of undeclared identifier '_dst'

The compiler output also shows:

253199 |     { uint32_t _src[4]$_dst[4]={0}; memcpy(_src,c->vreg[2],16); for(unsigned _j=0;_j<4;++_j) { uint64_t _v=_src[_j],_r=0;
                     ^
error: expected ';' at end of declaration

There are 3 errors total.

On nightly-287:

E:\OOT Remake\LongshotLauncher-data\work\sa\build\obj\code_026.split.c:261672:72: error: use of undeclared identifier 'Wm'
 261672 |     { uint32_t _n[4],_m[4],_r[4]={0}; memcpy(_n,c->vreg[0],16); memcpy(Wm,c->vreg[1],16);
        |                                                                        ^~

One error was reported. The nightly report's final 200 build-log lines and final 29 launcher-log lines are attached to the Discord report.

System and settings

Windows 10.0.26200 x64; .NET 10.0.15; 32 GB RAM, 24 threads. NVIDIA RTX 5070, Game Ready Driver 621.01. Free disk space: 575 GB. Build settings: memory and workers automatic, DLSS on, FSR 4 and XeSS off.

Logs

Launcher bug-report output includes build and launcher log excerpts. Still unknown: exact steps preceding the build attempts and whether other conditions affect reproduction.

Reported by [BobbyTortilla on Discord](https://discord.com/channels/1569849200047427645/1570784344241340417/1571684543541543033) ### What you used Longshot Launcher v1.0.0-rc.9 (portable), building with the Longshot builder. ### Version and channel Build v3.0.0-rc.4 (release candidate). The same failure happened on launcher rc.7 and rc.9. It also reproduces with Longshot nightly-287-0407ae24. ### Where it stopped Build stopped during Compile at 53% (`make_game.py` / `build.py`, exit status 1), while compiling `code_051.c` on the rc.4 attempt. On nightly-287 it stopped compiling `code_026.c` after 728 seconds total; that file took 7 seconds compiling alone. ### Error On rc.4: ```text E:\OOT Remake\LongshotLauncher-data\work\sa\build\obj\code_051.split.c:253203:51: error: use of undeclared identifier '_dst' ``` The compiler output also shows: ```text 253199 | { uint32_t _src[4]$_dst[4]={0}; memcpy(_src,c->vreg[2],16); for(unsigned _j=0;_j<4;++_j) { uint64_t _v=_src[_j],_r=0; ^ error: expected ';' at end of declaration ``` There are 3 errors total. On nightly-287: ```text E:\OOT Remake\LongshotLauncher-data\work\sa\build\obj\code_026.split.c:261672:72: error: use of undeclared identifier 'Wm' 261672 | { uint32_t _n[4],_m[4],_r[4]={0}; memcpy(_n,c->vreg[0],16); memcpy(Wm,c->vreg[1],16); | ^~ ``` One error was reported. The nightly report's final 200 build-log lines and final 29 launcher-log lines are attached to the Discord report. ### System and settings Windows 10.0.26200 x64; .NET 10.0.15; 32 GB RAM, 24 threads. NVIDIA RTX 5070, Game Ready Driver 621.01. Free disk space: 575 GB. Build settings: memory and workers automatic, DLSS on, FSR 4 and XeSS off. ### Logs Launcher bug-report output includes build and launcher log excerpts. Still unknown: exact steps preceding the build attempts and whether other conditions affect reproduction.
Author
Collaborator

Files from the Discord report:

Files from the Discord report: - [message.txt](https://git.magicbean.dev/attachments/7b95e631-0c7e-4e08-adb9-5fb34850640e)
Author
Collaborator

Updated with more from BobbyTortilla on Discord: Added the reporter's RTX 5070, Game Ready Driver 621.01, and 575 GB free disk space; the linked merged launcher PRs do not address this generated-C compiler failure.

Updated with more from [BobbyTortilla on Discord](https://discord.com/channels/1569849200047427645/1571684543541543033/1571688484408787027): Added the reporter's RTX 5070, Game Ready Driver 621.01, and 575 GB free disk space; the linked merged launcher PRs do not address this generated-C compiler failure.
Author
Collaborator

From BobbyTortilla on Discord:

BobbyTortilla reports the same compile failure on nightly: it stopped again at 53%, but this attempt reported undeclared identifier Wm instead. Exact nightly version and full compiler output for this second attempt were not provided. The builder-stop behavior fix in PR #540 is in nightly-285-a8a30c22, but it does not address the underlying generated-C compile error.

From [BobbyTortilla on Discord](https://discord.com/channels/1569849200047427645/1570784344241340417/1571684543541543033): BobbyTortilla reports the same compile failure on nightly: it stopped again at 53%, but this attempt reported undeclared identifier `Wm` instead. Exact nightly version and full compiler output for this second attempt were not provided. The builder-stop behavior fix in [PR #540](https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/540) is in nightly-285-a8a30c22, but it does not address the underlying generated-C compile error.
Author
Collaborator

Updated with more from BobbyTortilla on Discord: Added the exact nightly-287 version and its full Wm undeclared-identifier compiler error/log details.

Files:

Updated with more from [BobbyTortilla on Discord](https://discord.com/channels/1569849200047427645/1571684543541543033/1571696945804746914): Added the exact nightly-287 version and its full Wm undeclared-identifier compiler error/log details. Files: - [message.txt](https://git.magicbean.dev/attachments/2d6a9948-6d01-4a47-bd66-0b3e05b8fcbc)
Contributor

Both errors look like one flipped bit in the generated C, not a lift bug. The same bit both times:

  • rc.4, code_051.split.c: uint32_t _src[4]$_dst[4] should be _src[4],_dst[4]. ',' is 0x2C and '$' is 0x24.
  • nightly-287, code_026.split.c: memcpy(Wm,c->vreg[1],16) should be memcpy(_m,...). '_' is 0x5F and 'W' is 0x57. The same line spells _m correctly a few characters earlier (uint32_t _n[4],_m[4],...).

Each is a single byte with bit 3 (0x08) flipped, in a different file on a different build. The translator writes these lines the same way every time, and the same rc.4 and nightly sources build here and for other players. Random single-bit flips like that usually mean memory or CPU instability: an XMP/EXPO memory profile or an overclock that's slightly unstable, or a failing RAM stick. A long build that loads all 24 threads and lots of memory is often the first thing to show it.

Worth trying, in order:

  • turn off XMP/EXPO, or any overclock, in the BIOS, and build again;
  • run a memory test (Windows Memory Diagnostic, or MemTest86 for a few passes);
  • if a build gets further, a different file failing each time with a one-character typo confirms it.

Deleting work\sa\build\obj before retrying makes sure no corrupted object is reused.

Posted by an AI agent (Claude Code) for d4nkn3ss.

Both errors look like one flipped bit in the generated C, not a lift bug. The same bit both times: - rc.4, `code_051.split.c`: `uint32_t _src[4]$_dst[4]` should be `_src[4],_dst[4]`. ',' is 0x2C and '$' is 0x24. - nightly-287, `code_026.split.c`: `memcpy(Wm,c->vreg[1],16)` should be `memcpy(_m,...)`. '_' is 0x5F and 'W' is 0x57. The same line spells `_m` correctly a few characters earlier (`uint32_t _n[4],_m[4],...`). Each is a single byte with bit 3 (0x08) flipped, in a different file on a different build. The translator writes these lines the same way every time, and the same rc.4 and nightly sources build here and for other players. Random single-bit flips like that usually mean memory or CPU instability: an XMP/EXPO memory profile or an overclock that's slightly unstable, or a failing RAM stick. A long build that loads all 24 threads and lots of memory is often the first thing to show it. Worth trying, in order: - turn off XMP/EXPO, or any overclock, in the BIOS, and build again; - run a memory test (Windows Memory Diagnostic, or MemTest86 for a few passes); - if a build gets further, a different file failing each time with a one-character typo confirms it. Deleting `work\sa\build\obj` before retrying makes sure no corrupted object is reused. *Posted by an AI agent (Claude Code) for d4nkn3ss.*
Sign in to join this conversation.
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Longshot#985
No description provided.