Bug C backend: Convert Cptr to c_p_pointer

[ronnuriel]

Bug when converting Cptr to c_p_pointer (to work as array):

a: Pointer[u16[32768]] = c_p_pointer(vresult_l4, u16[32768])

code generation error: Type number '14' not supported

   --> /home/eamiel/tmp_cp_release_01/copperhead/demo-apps/belex-direct-fragment/device/../mat_vec_device.py:279:5

    |

279 |     a: Pointer[u16[32768]] = c_p_pointer(vresult_l4, u16[32768])

But i16 works:
a: Pointer[i16[32768]] = c_p_pointer(vresult_l4, i16[32768])

[certik]

Thanks, that's a great bug report. We'll fix it quickly.

[certik]

Looks like there are several related bugs, both in C and LLVM backends. The first minimal reproducible example to fix:

from lpython import CPtr, c_p_pointer, empty_c_void_p, Pointer, u16            

def test_tuple_return():
    e: CPtr = empty_c_void_p()
    iarray5: Pointer[u16[50]] = c_p_pointer(e, u16[50])         

test_tuple_return()

gives:

$ lpython bindc_07.py           
; ModuleID = 'LFortran'
source_filename = "LFortran"

%array = type { i16*, i32, %dimension_descriptor*, i1, i32 }
%dimension_descriptor = type { i32, i32, i32 }

define void @__module__global_symbols__lpython_main_program() {
.entry:
  call void @__module__global_symbols_test_tuple_return()
  br label %return

return:                                           ; preds = %.entry
  ret void
}

define void @__module__global_symbols_test_tuple_return() {
.entry:
  %e = alloca void*, align 8
  %iarray5 = alloca %array*, align 8
  store void* null, void** %e, align 8
  %0 = load void*, void** %e, align 8
  %1 = alloca <1 x i32>, align 4
  %2 = getelementptr <1 x i32>, <1 x i32>* %1, i32 0, i32 0
  store i32 50, i32* %2, align 4
  %3 = getelementptr <1 x i32>, <1 x i32>* %1, i32 0, i32 0
  %4 = alloca %array, align 8
  %5 = getelementptr %array, %array* %4, i32 0, i32 1
  store i32 0, i32* %5, align 4
  %6 = alloca %dimension_descriptor, align 8
  %7 = getelementptr %array, %array* %4, i32 0, i32 2
  store %dimension_descriptor* %6, %dimension_descriptor** %7, align 8
  %8 = getelementptr %array, %array* %4, i32 0, i32 4
  store i32 1, i32* %8, align 4
  store %array* %4, void* %0, align 8
  %9 = getelementptr %array, %array* %4, i32 0, i32 0
  %10 = getelementptr %array, %array* %4, i32 0, i32 2
  %11 = load %dimension_descriptor*, %dimension_descriptor** %10, align 8
  %12 = bitcast void* %0 to i16*
  store i16* %12, i16** %9, align 8
  %13 = getelementptr inbounds %dimension_descriptor, %dimension_descriptor* %11, i32 0
  %14 = getelementptr %dimension_descriptor, %dimension_descriptor* %13, i32 0, i32 0
  %15 = getelementptr %dimension_descriptor, %dimension_descriptor* %13, i32 0, i32 1
  %16 = getelementptr inbounds %dimension_descriptor, %dimension_descriptor* %11, i32 0
  %17 = getelementptr %dimension_descriptor, %dimension_descriptor* %16, i32 0, i32 2
  store i32 1, i32* %14, align 4
  %18 = getelementptr inbounds i32, i32* %3, i32 0
  %19 = load i32, i32* %18, align 4
  store i32 0, i32* %15, align 4
  %20 = sub i32 %19, 0
  %21 = add i32 %20, 1
  store i32 %21, i32* %17, align 4
  %22 = mul i32 1, %21
  br label %return

return:                                           ; preds = %.entry
  ret void
}

define i32 @main(i32 %0, i8** %1) {
.entry:
  call void @_lpython_set_argv(i32 %0, i8** %1)
  call void @__module__global_symbols__lpython_main_program()
  ret i32 0
}

declare void @_lpython_set_argv(i32, i8**)
code generation error: asr_to_llvm: module failed verification. Error:
Stored value type does not match pointer operand type!
  store %array* %4, void* %0, align 8
 void


Note: if any of the above error or warning messages are not clear or are lacking
context please report it to us (we consider that a bug that must be fixed).

[certik]

@ronnuriel the issue should be now fixed, please go ahead and try the latest LPython version and see if it works.