What causes "could not set register %s: not found" in go-delve/delve?

Verify the register name against the stub's target.xml (fetch qXfer:features:read) or qRegisterInfo output

could not set register %s: not found

Error message

could not set register %s: not found

What it means

SetReg looks up the register name in the thread's register map (t.regs.regs, populated from target.xml/qRegisterInfo at handshake). If the name is absent, the register cannot be written and this error is returned.

Source

Thrown at pkg/proc/gdbserial/gdbserver.go:2099

		gdbreg, ok = t.regs.regs["y"+regName[1:]]
		if !ok {
			gdbreg, ok = t.regs.regs["z"+regName[1:]]
		}
	}
	if !ok && t.p.bi.Arch.Name == "arm64" && regName == "x30" {
		gdbreg, ok = t.regs.regs["lr"]
	}
	if !ok && regName == "rflags" {
		// rr has eflags instead of rflags
		regName = "eflags"
		gdbreg, ok = t.regs.regs[regName]
		if ok {
			reg.FillBytes()
			reg.Bytes = reg.Bytes[:4]
		}
	}
	if !ok {
		return fmt.Errorf("could not set register %s: not found", regName)
	}
	reg.FillBytes()

	wrongSizeErr := func(n int) error {
		return fmt.Errorf("could not set register %s: wrong size, expected %d got %d", regName, n, len(reg.Bytes))
	}

	if len(reg.Bytes) == len(gdbreg.value) {
		copy(gdbreg.value, reg.Bytes)
		err := t.p.conn.writeRegister(t.strID, gdbreg.regnum, gdbreg.value)
		if err != nil {
			return err
		}
		if t.p.conn.workaroundReg != nil && len(gdbreg.value) > 16 {
			// This is a workaround for a bug in debugserver where register writes (P
			// packet) on AVX-2 and AVX-512 registers are ignored unless they are
			// followed by a write to an AVX register.
			// See:

View on GitHub (pinned to a23773e6c3)

Solutions

Verify the register name against the stub's target.xml (fetch qXfer:features:read) or qRegisterInfo output
Use a stub with a complete target.xml for your architecture (recent gdbserver)
Do not attempt to write registers unsupported by the stub; guard code by checking availability first
Update Delve if the register-name mapping for your architecture is outdated

Defensive patterns

Strategy: validation

Validate before calling

// check the register exists before writing:
_, err := thread.Registers()
hasReg := func(regs proc.Registers, name string) bool {
    for _, r := range regs.Slice(false) { if r.Name == name { return true } }
    return false
}

Try / catch

err := thread.SetReg(reg)
if err != nil && strings.Contains(err.Error(), "not found") {
    return fmt.Errorf("stub does not expose register %q: %w", reg.Name, err)
}

Prevention

Fetch and inspect the stub's target.xml to know the writable register set
Match stub architecture to the inferior architecture
Skip writes for registers absent from the stub's register list
Use current gdbserver/lldb-server builds with full register descriptions

When it happens

Trigger: Writing a register whose name is not exposed by the stub's target.xml or qRegisterInfo — e.g. writing an architecture-specific register (x87/SSE/debug regs) on a stub that does not list it, or a typo'd/renamed register between architectures.

Common situations: Using a stub that exposes a reduced register set (minimal target.xml); mixed-architecture remote debugging where the register name differs (e.g. rip vs pc); stubs built without FPU/debug-register descriptions.

Related errors

AI-assisted analysis of go-delve/delve@a23773e6c3 (2026-08-31). Data as JSON: /api/errors/ad2b2a80a7e68aae. Report an issue: GitHub.