fix alignment of emit_unbox_store copy

The dest alignment might be determined to be greater than the source.
For example, observed with:

    code_llvm(dump_module=true, optimize=false, (Int,)) do x
        Pair(ntuple(i -> 0x00, 8), x)
    end

Where the alignment of the first field of the Pair is at least 4, but
the alignment of the ntuple data is 1.

doc/src/devdocs/debuggingtips.md

@@ -41,11 +41,16 @@ useful.
 
 ## Useful Julia functions for Inspecting those variables
 
-  * `jl_gdblookup($rip)` :: For looking up the current function and line. (use `$eip` on i686 platforms)
+  * `jl_print_task_backtraces(0)` :: Similar to gdb's `thread apply all bt` or lldb's `thread backtrace
+    all`. Runs all threads while printing backtraces for all existing tasks.
+  * `jl_gdblookup($pc)` :: For looking up the current function and line.
+  * `jl_gdblookupinfo($pc)` :: For looking up the current method instance object.
+  * `jl_gdbdumpcode(mi)` :: For dumping all of `code_typed/code_llvm/code_asm` when the REPL is not working right.
   * `jlbacktrace()` :: For dumping the current Julia backtrace stack to stderr. Only usable after
     `record_backtrace()` has been called.
   * `jl_dump_llvm_value(Value*)` :: For invoking `Value->dump()` in gdb, where it doesn't work natively.
     For example, `f->linfo->functionObject`, `f->linfo->specFunctionObject`, and `to_function(f->linfo)`.
+  * `jl_dump_llvm_module(Module*)` :: For invoking `Module->dump()` in gdb, where it doesn't work natively.
   * `Type->dump()` :: only works in lldb. Note: add something like `;1` to prevent lldb from printing
     its prompt over the output
   * `jl_eval_string("expr")` :: for invoking side-effects to modify the current state or to lookup

src/aotcompile.cpp

@@ -1820,6 +1820,58 @@ void addTargetPasses(legacy::PassManagerBase *PM, const Triple &triple, TargetIR
     PM->add(createTargetTransformInfoWrapperPass(std::move(analysis)));
 }
 
+// sometimes in GDB you want to find out what code was created from a mi
+extern "C" JL_DLLEXPORT_CODEGEN jl_code_info_t *jl_gdbdumpcode(jl_method_instance_t *mi)
+{
+    jl_llvmf_dump_t llvmf_dump;
+    size_t world = jl_current_task->world_age;
+    JL_STREAM *stream = (JL_STREAM*)STDERR_FILENO;
+
+    jl_printf(stream, "---- dumping IR for ----\n");
+    jl_static_show(stream, (jl_value_t*)mi);
+    jl_printf(stream, "\n----\n");
+
+    jl_printf(stream, "\n---- unoptimized IR ----");
+    jl_get_llvmf_defn(&llvmf_dump, mi, world, 0, false, jl_default_cgparams);
+    if (llvmf_dump.F) {
+        jl_value_t *ir = jl_dump_function_ir(&llvmf_dump, 0, 1, "source");
+        jl_static_show(stream, ir);
+    }
+    jl_printf(stream, "----\n");
+
+    jl_printf(stream, "\n---- optimized IR ----");
+    jl_get_llvmf_defn(&llvmf_dump, mi, world, 0, true, jl_default_cgparams);
+    if (llvmf_dump.F) {
+        jl_value_t *ir = jl_dump_function_ir(&llvmf_dump, 0, 1, "source");
+        jl_static_show(stream, ir);
+    }
+    jl_printf(stream, "----\n");
+
+    jl_printf(stream, "\n---- assembly ----");
+    jl_get_llvmf_defn(&llvmf_dump, mi, world, 0, true, jl_default_cgparams);
+    if (llvmf_dump.F) {
+        jl_value_t *ir = jl_dump_function_asm(&llvmf_dump, 0, "", "source", 0, true);
+        jl_static_show(stream, ir);
+    }
+    jl_printf(stream, "----\n");
+
+    jl_code_info_t *src = NULL;
+    jl_value_t *ci = jl_default_cgparams.lookup(mi, world, world);
+    if (ci != jl_nothing) {
+        jl_code_instance_t *codeinst = (jl_code_instance_t*)ci;
+        src = (jl_code_info_t*)jl_atomic_load_relaxed(&codeinst->inferred);
+        if ((jl_value_t*)src != jl_nothing && !jl_is_code_info(src) && jl_is_method(mi->def.method)) {
+            JL_GC_PUSH1(&codeinst);
+            src = jl_uncompress_ir(mi->def.method, codeinst, (jl_value_t*)src);
+            JL_GC_POP();
+        }
+    }
+    if (!src || (jl_value_t*)src == jl_nothing) {
+        src = jl_type_infer(mi, world, 0);
+    }
+    return src;
+}
+
 // --- native code info, and dump function to IR and ASM ---
 // Get pointer to llvm::Function instance, compiling if necessary
 // for use in reflection from Julia.

src/cgutils.cpp

@@ -2911,7 +2911,8 @@ static void init_bits_cgval(jl_codectx_t &ctx, Value *newv, const jl_cgval_t& v,
 {
     // newv should already be tagged
     if (v.ispointer()) {
-        emit_memcpy(ctx, newv, jl_aliasinfo_t::fromTBAA(ctx, tbaa), v, jl_datatype_size(v.typ), sizeof(void*), julia_alignment(v.typ));
+        unsigned align = std::max(julia_alignment(v.typ), (unsigned)sizeof(void*));
+        emit_memcpy(ctx, newv, jl_aliasinfo_t::fromTBAA(ctx, tbaa), v, jl_datatype_size(v.typ), align, julia_alignment(v.typ));
     }
     else {
         init_bits_value(ctx, newv, v.V, tbaa);

src/intrinsics.cpp

@@ -520,7 +520,7 @@ static void emit_unbox_store(jl_codectx_t &ctx, const jl_cgval_t &x, Value *dest
     }
 
     Value *src = data_pointer(ctx, x);
-    emit_memcpy(ctx, dest, jl_aliasinfo_t::fromTBAA(ctx, tbaa_dest), src, jl_aliasinfo_t::fromTBAA(ctx, x.tbaa), jl_datatype_size(x.typ), alignment, alignment, isVolatile);
+    emit_memcpy(ctx, dest, jl_aliasinfo_t::fromTBAA(ctx, tbaa_dest), src, jl_aliasinfo_t::fromTBAA(ctx, x.tbaa), jl_datatype_size(x.typ), alignment, julia_alignment(x.typ), isVolatile);
 }
 
 static jl_datatype_t *staticeval_bitstype(const jl_cgval_t &targ)