<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/202186>202186</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            [RISC-V][lld] ld.lld big endian emulation and output format names unsupported
        </td>
    </tr>

    <tr>
      <th>Labels</th>
      <td>
            lld
      </td>
    </tr>

    <tr>
      <th>Assignees</th>
      <td>
      </td>
    </tr>

    <tr>
      <th>Reporter</th>
      <td>
          XiaobingHou1219
      </td>
    </tr>
</table>

<pre>
    I ran into this while reducing a RISC-V linker testcase. The reproducer is small, and I have been seeing the same result consistently across three reruns.

### Summary
ld.lld 22.1.0 links the big-endian objects successfully when the target is inferred from the input object, but rejects explicit big-endian selectors: `ld.lld -melf32briscv` and `-melf64briscv` fail with `unknown emulation`, and linker scripts containing `OUTPUT_FORMAT(elf32-bigriscv)` or `OUTPUT_FORMAT(elf64-bigriscv)` fail with `unknown output format name`. GNU ld.bfd 2.45 accepts the same explicit emulations and output-format scripts and emits ELF headers marked big-endian.

### Expected behavior
Link LLVM-MC-generated big-endian RISC-V objects by explicitly selecting the GNU-compatible big-endian target through `ld.lld -melf32briscv`/`-melf64briscv` or through linker-script `OUTPUT_FORMAT(elf32-bigriscv)`/`OUTPUT_FORMAT(elf64-bigriscv)`. Since ld.lld already accepts the same input objects by target inference and emits big-endian RISC-V ELF outputs, the explicit selector routes should be recognized consistently rather than rejected before normal link processing.

### Environment
- linker route: LLVM MC 22 `riscv32be`/`riscv64be` objects plus current ld.lld 22.1.0; current GNU ld.bfd 2.45 with the same `-melf32briscv`/`-melf64briscv` and `OUTPUT_FORMAT` scripts as positive controls
- march: riscv32be and riscv64be
- mabi: ELF32/ELF64 big-endian RISC-V objects from llvm-mc
- first failing stage: link
- local stability check: True

### Reduced testcase
These are the reduced input files from the minimized reproducer I used locally:

#### `case.s`
```asm
.globl _start
_start:
 ret
```

### Reproduction notes
- This packaged root does not have a single canonical `run.ps1` wrapper.
- Use the reduced inputs under `case/` and follow the commands documented in `case/README.md`.
- Stable witness outputs, when present, are preserved under `verify/run1..run3/`.

### What I checked
- Reduced inputs are preserved under case/.
- Stable witness outputs are preserved under verify/run1..run3/.
- The strict recheck says stable normalized run signatures across three runs: True.
- Tracker guidance link: https://llvm.org/docs/HowToSubmitABug.html
- evidence summary: 3 clean reproductions under hunt/verify/lld_be_emulation_names/run1..run3` show stable LLVM-MC assembly success, stable lld inference success, stable lld explicit-selector rejection, and stable GNU ld.bfd success. The combined summary has 36 rows: `lld_infer_success=6`, `lld_emulation_unknown=6`, `lld_output_format_unknown=6`, `bfd_emulation_success=6`, and `bfd_output_format_success=6`. Representative run1 RV64 `lld-infer` output has a big-endian ELF header, while run1 `lld-emulation` stderr is `ld.lld: error: unknown emulation: elf64briscv` and run1 `lld-output-format` stderr is `ld.lld: error: unknown output format name: elf64-bigriscv`. Local source/test search shows `upstream/llvm-project/lld/test/ELF/emulation-riscv.s` covering only little-endian `elf32lriscv`/`elf64lriscv` and `elf32-littleriscv`/`elf64-littleriscv`. Web searches on 2026-06-05 over LLVM GitHub issues/PRs and broad web queries for `elf32briscv`, `elf64briscv`, `unknown emulation: elf64briscv`, and `elf64-bigriscv ld.lld` found no direct current upstream issue report; results were generic LLVM issues or unrelated RISC-V compiler/linker reports.

### Notes
upstream/llvm-project/lld/test/ELF/emulation-riscv.s`. GNU ld.bfd 2.45 accepts `elf32briscv`/`elf64briscv` and `elf32-bigriscv`/`elf64-bigriscv` locally. The RISC-V psABI currently notes that the big-endian ABI is not yet included, so this should be reported as ld.lld self-consistency/GNU-compatibility for its implemented `elf*-bigriscv` output support rather than ratified psABI conformance.

Root key: `lld.riscv.big_endian_emulation_and_output_format_names_unsupported`
Case id: `20260605-lld-riscv-big-endian-emulation-output-format-unsupported`

</pre>
<img width="1" height="1" alt="" src="http://email.email.llvm.org/o/eJykWEtv4zgS_jXKpWBBkW0lOeSQl7sDpHsG6aRnbwFFli1uKNLLRzzeX78okpLtxJnpxQBBw00Wi_X4quqjmHNypREvi_l1Mb89YcF3xl7-SzLTSr36asJpfXpx0hqxvbwHyzRI7Q34TjrYdFIhWBSBS70CBo_3P24mP0FJ_YoWPDrPmcMSnjoSW1sjAkcL0oHrmVJFfQNMC7iHjr0htIgaHCLp8h2CYz0dc0F54EY76Txqr7bAuDXOge8skoAN2pVFdVtUV_nfepr-4Efoe2a3aVmJUikBdV2ellU00sV7WrmaoBaSaTDtv5F7By5wjs4tg1Jb2HSoo6BndoWezJd6idaigKU1fdyTeh18Pk9-tcGDxaQN_1wryaXfv8mhQu6NdcX0CoqmysZNelTLad1a6fhb0VQxPkVTxfVmtltfMqlgI31Hu0G_arPRgH1QzEuji6YagpuT4biVa-8okJ5JTTEumuq356ffn59eFr89frt6KurzePmklat0T31BVxn7iWgzey961CoTPMVmaWzPPGjWY9FUJXz5_gxKlO1SQF3O5sA4R7JwTP0YttEtFz1K-iZZ3-AY7WAvvYO7hwV0yARaBz2zryj2Av8JUO7-XCP3JIkde5PGJoEHqV_h4eHnt8m3m8kKNVrmD9QNkB-A025Hs9U2J3nA85fvzxNu-jXzslUHsMvA8p01YdX9BRyKenEMDMaOZ1O6Jykqv5ripPZXMlzCD6k5QraPKYtMbD_mbr8cYlSG2qHCQdKwy9fHaFIGU5YdwZi0jmAYCgesCR4duM4ERXkDi9ystPwvisN2YZnvqB11TOeSjHleGougCUQqRg3W1lDRS736DCT6TVqje9Q-bU2G6oq2UCUTVuDbDdQ1xT4Gblq3OMY4rjSzuDKGZ62CAx6sRe3hoEsV0-tx4329xDIbAz7A4u_BkjvKYbabaldIDtbGSS_fMHYLa5Qb3O2Z5R35OXoW1e28GuVaSWJ3D4tpXdSLu4dFM_uLsoltVKm3ftLzQcdSWudjR6ECcp6tYoQp4mP0DWeKtlqppN8C75C_ktCTDXg8h480rVCMwyntP3XoEJjFGFCbZRKIl1Kh2zX6XmrZR5DtTbR7CA5Fskdti-nV0cvp_qKp4kx0lKC03VTpj7k-rZQrZVoFL84zm5GWfw-KwaJ_d_ozd5ON1D1BG49jKp9ogq8Zf2Ur8sUYD8KgI6E0jhlQKSgEzrTRkiJNAA66XLtTAszGsvUabTlofHZHwucgaIF28DthMoJmaZQym3iCm75nWjgQhgcqr3h678zj3dXtt7uyF9SChvt-eEZ9dCO9Ruf2G0ac2GuLjiqV5qDF9F_7hmJn0BtaudwW9cIGfVqWNuhpsu-T-v-jYx7uE8pQDGY8Hrp77K7sxd9YfvTocRvLXRYRnLeSE9uIhoFjWxdrQg3tLaE1aCCqx3yw6N5xqKDdUDY71ZZx6m2rIAWLPZ8qb3oFnfdrEqdY1Quq2tLYVVEvhOGuqBdfzebJ_AhtL_3VdViVne_VoBTfpIjt32VmNr2CKXCFsTfvsDrApguUwcUYBaXES4svIyN4ITrhDsNDvawzmyEGeXwDcw77lqZyIncEjCxCDXc3mY7vDwNoshtAcZYQ28pUK0vvNeqsKjFgbvpWahSD79AxB9MGrNmMLFCJl2jIy2DE9LbJZC5v71zP_OqjSMLTSyJIx8Xa5b6mj5flIUFih9oORcvYYGKdsTgwKA3w-LOZZVsm0Zs47BILJJ_Z_iTYsbVUufFJQVqygn1OC84LtPEBMXIkihxaayz9-EiEaffj_Nu_4IBO_j-XHOG1w2071kQxekhjygTLqQ3Q6AGHNEojUOM9Ye28RdbnipqsrUlPCYJ8PpQGaVEvRvcm8ZI4TIAbqhK9AqPVFpT0XuEQ5KKpIjdQh9wg2qreM4PEEJOCI_Lvdkr4A9vsDjowGuqqbiZVM6nmQCYlTvRF-q-hBelciPX6-2Oi7K01TMAGW_hPQCtp0qb3xnsucwODAe8XfyHpe5A-zE-mW_HtYoIWoA0IaZH7kXcNmUm2U5cyNImv89PUwQYtQnwcSJ6cTV4SLQ_aoopvhsx46AEgFYF9MXDHqPCzF-z33cz-xxD5_M11JN6LD-E-xMcewvfBsbc80KHU_bL_a3d1fT_EVm0TJyFu7t-_xUlOJj6yjS8HroJAEdty_gCxz_0piCiIvWYG7VAtJ-NDgNP02H-BJcZIYKM3iOzXCjPzSL4U9dWBM7naXVjTTYePCublUqIYnDM69gTN8SCpj8SxXnG76_Vlyk0rVy_J6b2ezPT71htH3UvQ2QQUI-u7YQ5BiqyY6q9qqvmEulu8YbIL666fHja-yTG9J-JyKi6mF-wEL0_PzqvzaXMxb066y3rWiHMxP6_auhXnDYrlDM_Oz2ZVPZ8u61l1Ii-zFWen9Xw-b8qL5qw-b2s-5-L8jDfTYlZhz6QqB_5wEouGjp2eNyeKtahc_ChV1xHedTG_PbGXEfltWLliVinpvNsp8NKr-B0rQa2Y3xbzazo7vx0w0coVZHiNcdj7prDfy4mBjBE5CVZdHvKelfRdaEtuhnr8WJZjs8tOvV3W_wsAAP__pcF2Yg">