[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu.org

--- Comment #22 from Tobias Burnus <burnus at gcc dot gnu.org> 2010-09-28 20:43:57 UTC ---
Another case of off-by-one line numbers:
  http://gcc.gnu.org/ml/fortran/2010-09/msg00500.html

For
     if (Gridded_resid_mask(4,4)) then  ! Line 5
          write(*,*)  'Boo'             ! Line 6

The "if" is in the wrong line (6 instead of 5) - while the condition itself is
in the correct line; i.e.:

  [ghghjh.f90 : 6] if ([ghghjh.f90 : 5] (*gridded_resid_mask)[[ghghjh.f90 : 5]
(stride.2 + 1) * 4 + offset.3])
    {
[...]
        [ghghjh.f90 : 6] _gfortran_transfer_character ([ghghjh.f90 : 6]
&dt_parm.7, [ghghjh.f90 : 6] &[ghghjh.f90 : 6] "Boo"[1]{lb: 1 sz: 1}, 3);

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

--- Comment #23 from Tobias Burnus <burnus at gcc dot gnu.org> 2010-10-15 12:42:43 UTC ---
Author: burnus
Date: Fri Oct 15 12:42:39 2010
New Revision: 165507

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=165507
Log:
2010-10-15  Tobias Burnus  <burnus@net-b.de>

        PR fortran/45186
        * trans.h (gfc_add_modify_loc, gfc_evaluate_now_loc): New
        * prototypes.
        (gfc_trans_runtime_error_vararg): Remove prototype.
        * trans.c (gfc_add_modify_loc, gfc_evaluate_now_loc): New
        * functions.
        (gfc_add_modify, gfc_evaluate_now): Use them.
        (trans_runtime_error_vararg): Renamed from
        gfc_trans_runtime_error_vararg, made static and use locus.
        (gfc_trans_runtime_error): Use it.
        (gfc_trans_runtime_check): Ditto and make use of locus.
        * trans-stmt.c (gfc_trans_if_1, gfc_trans_simple_do,
        gfc_trans_do, gfc_trans_do_while): Improve line number
        associated with generated expressions.


Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/trans-stmt.c
    trunk/gcc/fortran/trans.c
    trunk/gcc/fortran/trans.h

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #24 from Tobias Burnus <burnus at gcc dot gnu.org> 2010-10-15 12:45:59 UTC ---
FIXED on the trunk (4.6).

The last commit now also made the debug lines for DO, DO WHILE and IF blocks
more helpful. Thus, I am closing this PR.

I am sure that there are still issues with debug lines. Please fill a PR if you
find them!

And thanks again, Nikolay, for reporting the bug.

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #21 from burnus at gcc dot gnu dot org  2010-09-10 07:25 -------
(In reply to comment #20)
> many thanks for working on this... I mentioned this before in the PR, but it
> would be very good if some line number testcases were added to the regression
> tests.

I concur - with -dump-tree-original-lineno this should be possible.

> If there would be some semi-automatic way of checking the lineno quality (such
> as e.g. in the original dump line numbers should always increase??)

No! As the example in comment 18 shows the line numbers should not always
monotonically  increase. Or at least I expect the "i = i + 1" to be associated
with the line
   do i = 1, 10
rather than with the line
      a(i) = b(i)
as the gfortran currently does. If you disagree, I can simply close this PR as
fixed ;-)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[jv244 at cam dot ac dot uk]

------- Comment #20 from jv244 at cam dot ac dot uk  2010-09-10 06:36 -------
Tobias,

many thanks for working on this... I mentioned this before in the PR, but it
would be very good if some line number testcases were added to the regression
tests. Both for performance measurements and debugging, this is essential info,
and it is too remarkable good lineno info could 'disappear' from a released
(4.5) compiler just like that.

If there would be some semi-automatic way of checking the lineno quality (such
as e.g. in the original dump line numbers should always increase??), I'd be
happy to give it a try on CP2K. 

Joost


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #19 from burnus at gcc dot gnu dot org  2010-09-10 06:02 -------
Subject: Bug 45186

Author: burnus
Date: Fri Sep 10 06:01:50 2010
New Revision: 164143

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=164143
Log:
2010-09-10  Tobias Burnus  <burnus@net-b.de>

        PR fortran/45186
        * trans.h (build1_stat_loc, build2_stat_loc, build3_stat_loc,
        build4_stat_loc): New inline functions.
        (build1_loc, build2_loc, build3_loc, build4_loc): New macros.
        (build1_v, build2_v, build3_v, build4_v): Use input_location
        as locus.
        * trans-array.c (gfc_trans_scalarized_loop_end,
        gfc_conv_array_parameter): Replace build[1-4] by build[1-4]_loc.
        * trans.c (gfc_build_addr_expr, gfc_build_array_ref,
        gfc_finish_wrapped_block): Ditto.
        * trans-decl.c (gfc_init_default_dt, init_intent_out_dt): Ditto.
        * trans-expr.c (gfc_conv_missing_dummy,
        gfc_trans_alloc_subarray_assign, gfc_trans_zero_assign): Ditto.
        * trans-openmp.c (gfc_omp_clause_default_ctor,
        gfc_trans_omp_critical, gfc_trans_omp_parallel,
        gfc_trans_omp_parallel_do, gfc_trans_omp_parallel_sections,
        gfc_trans_omp_parallel_workshare, gfc_trans_omp_sections
        gfc_trans_omp_single, gfc_trans_omp_task,
        gfc_trans_omp_workshare): Ditto.


Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/trans-array.c
    trunk/gcc/fortran/trans-decl.c
    trunk/gcc/fortran/trans-expr.c
    trunk/gcc/fortran/trans-openmp.c
    trunk/gcc/fortran/trans.c
    trunk/gcc/fortran/trans.h


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #18 from burnus at gcc dot gnu dot org  2010-09-09 19:33 -------
(In reply to comment #0)
> This is pretty widespread. I have attached one example

That's now mostly fixed - though one patch is pending. Now
  gfortran -fdump-tree-original-lineno
generates mostly sensible line numbers.

For the given example, only for the DO loop the line numbers are a bit counter
intuitive, given the line numbers
    14    do i = 1,10
    15       array1(i) = i
    16    end do

the dump of the internal representation (shown in a C-like syntax) is:

  [debug.f90 : 14] i = 1;
  [debug.f90 : 15] if ([debug.f90 : 15] i <= 10)
...
         [debug.f90 : 15] array1[[debug.f90 : 15] (integer(kind=8)) i + -1] =
i;
         L.1:;
         [debug.f90 : 15] D.1577 = [debug.f90 : 15] i == 10;
         [debug.f90 : 15] i = [debug.f90 : 15] i + 1;
         [debug.f90 : 15] if (D.1577) goto L.2;

The "i = 1" is in the expected line - as is the "array1(i) = i" assignment, but
the loop increment and the exit conditions I would expect in line 14 and not in
line 15.

If no one else does, I will have a look after the last "input_location" patch
has been committed. I expect that there are more cases where the annotation is
slightly misplaced.


If one looks at GCC 4.4, one finds:
  [debug.f90 : 14] i = 1;
  [debug.f90 : 14] if (i <= 10)
         [debug.f90 : 15] array1[(integer(kind=8)) i + -1] = i;
         L.1:;
         D.1550 = i == 10;
         i = i + 1;
         if (D.1550) goto L.2;

Thus, the first loop condition is in the proper line (14) while there is no
line number associated with the loop increment and the exit condition.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #17 from burnus at gcc dot gnu dot org  2010-09-07 19:03 -------
Subject: Bug 45186

Author: burnus
Date: Tue Sep  7 19:03:41 2010
New Revision: 163964

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=163964
Log:
2010-09-07  Tobias Burnus <burnus@net-b.de>

        PR fortran/45186
        * trans-array.c (gfc_conv_descriptor_data_get,
        gfc_conv_descriptor_data_set, gfc_conv_descriptor_data_addr,
        gfc_conv_descriptor_offset, gfc_conv_descriptor_dtype,
        gfc_conv_descriptor_dimension, gfc_conv_descriptor_stride,
        gfc_conv_descriptor_lbound, gfc_conv_descriptor_ubound,
        gfc_conv_shift_descriptor_lbound,
        gfc_set_loop_bounds_from_array_spec,
        gfc_trans_allocate_array_storage, gfc_trans_create_temp_array,
        gfc_conv_array_transpose, gfc_get_iteration_count,
        gfc_grow_array, gfc_trans_array_ctor_element,
        gfc_trans_array_constructor_subarray,
        gfc_trans_array_constructor_value,
        constant_array_constructor_loop_size,
        gfc_trans_array_constructor,
        gfc_set_vector_loop_bounds, gfc_trans_array_bound_check,
        gfc_conv_array_index_offset, gfc_conv_scalarized_array_ref,
        gfc_conv_array_ref, gfc_trans_preloop_setup,
        gfc_trans_scalarized_loop_end, gfc_conv_ss_startstride,
        gfc_conv_loop_setup, gfc_conv_array_extent_dim,
        gfc_conv_descriptor_size, gfc_array_init_size,
        gfc_array_allocate, gfc_array_deallocate,
        gfc_trans_array_bounds, gfc_trans_auto_array_allocation,
        gfc_trans_dummy_array_bias, gfc_get_dataptr_offset,
        get_array_charlen, gfc_conv_expr_descriptor,
        array_parameter_size, gfc_conv_array_parameter,
        gfc_trans_dealloc_allocated, get_full_array_size,
        duplicate_allocatable,
        structure_alloc_comps): Change fold_build[0-9] to
        fold_build[0-9]_loc.
        (duplicate_allocatable, structure_alloc_comps,
        gfc_duplicate_allocatable): Add space after function name.


Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/trans-array.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #16 from burnus at gcc dot gnu dot org  2010-09-07 15:29 -------
(In reply to comment #15)
> Created an attachment (id=21725)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=21725&action=view) [edit]

+#define build6_loc(l,c,t1,t2,t3,t4,t5,t6,t7) \
+  build1_stat_loc (l,c,t1,t2,t3,t4,t5,t6,t7 MEM_STAT_INFO)

The last line should use -6- and not -1-, i.e. build1_stat_loc ->
build6_stat_loc


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #15 from burnus at gcc dot gnu dot org  2010-09-07 11:37 -------
Created an attachment (id=21725)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=21725&action=view)
tree.h patch tp add build[0-9]_loc

(In reply to comment #9)
> grep -w build[0-9] trans*.c | wc -l
> 38
> grep -w build[0-9]_v trans*.c | wc -l
> 158

That part requires the attached patch to add build[0-9]_loc to gcc/tree.h. The
patch still needs to be build with all languages - and then "make -k check"
regtested ...

> grep -w fold_build[0-9] trans*.c | wc -l
> 734

That's now down to
  trans-array.c:198
and down to 117 in my local tree.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #14 from burnus at gcc dot gnu dot org  2010-09-05 18:10 -------
Subject: Bug 45186

Author: burnus
Date: Sun Sep  5 18:10:11 2010
New Revision: 163879

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=163879
Log:
2010-09-05  Tobias Burnus  <burnus@net-b.de>

        PR fortran/45186
        * trans-intrinsic.c (gfc_conv_intrinsic_sign,
        gfc_conv_intrinsic_leadz): Use build_call_expr_loc instead
        of build_call_expr.
        * trans-expr.c (gfc_conv_expr_present, gfc_conv_missing_dummy,
        gfc_conv_string_length, gfc_conv_substring,
        gfc_conv_component_ref, gfc_conv_unary_op, gfc_conv_powi,
        gfc_conv_cst_int_power, gfc_conv_string_tmp, gfc_conv_concat_op,
        gfc_conv_expr_op, gfc_build_compare_string,
        gfc_set_interface_mapping_bounds, gfc_conv_subref_array_arg,
        gfc_conv_derived_to_class, conv_isocbinding_procedure,
        gfc_conv_procedure_call, fill_with_spaces,
        gfc_trans_string_copy, gfc_trans_alloc_subarray_assign,
        gfc_trans_structure_assign, gfc_trans_pointer_assignment,
        gfc_trans_scalar_assign, gfc_trans_zero_assign,
        gfc_trans_array_copy, gfc_trans_array_constructor_copy): Change
        fold_build[0-9] to fold_build[0-9]_loc.
        * trans-io.c (set_parameter_const, set_parameter_value,
        set_parameter_ref, gfc_convert_array_to_string, set_string,
        set_internal_unit, io_result, set_error_locus,
        nml_get_addr_expr, build_dt): Ditto.
        * trans-openmp.c (gfc_omp_clause_default_ctor,
        gfc_omp_clause_copy_ctor, gfc_omp_clause_assign_op,
        gfc_trans_omp_array_reduction, gfc_trans_omp_atomic,
        gfc_trans_omp_do): Ditto.
        * trans.c (gfc_add_modify, gfc_build_addr_expr,
        gfc_build_array_ref, gfc_trans_runtime_error_vararg,
        gfc_trans_runtime_check, gfc_call_malloc,
        gfc_allocate_with_status, gfc_allocate_array_with_status,
        gfc_call_free, gfc_deallocate_with_status,
        gfc_call_realloc): Ditto.


Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/convert.c
    trunk/gcc/fortran/f95-lang.c
    trunk/gcc/fortran/trans-intrinsic.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #13 from burnus at gcc dot gnu dot org  2010-09-03 19:41 -------
Subject: Bug 45186

Author: burnus
Date: Fri Sep  3 19:41:11 2010
New Revision: 163838

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=163838
Log:
2010-09-03  Tobias Burnus  <burnus@net-b.de>

        PR fortran/45186
        * trans-intrinsic.c (gfc_conv_intrinsic_sign,
        gfc_conv_intrinsic_leadz): Use build_call_expr_loc instead
        of build_call_expr.
        * trans-expr.c (gfc_conv_expr_present, gfc_conv_missing_dummy,
        gfc_conv_string_length, gfc_conv_substring,
        gfc_conv_component_ref, gfc_conv_unary_op, gfc_conv_powi,
        gfc_conv_cst_int_power, gfc_conv_string_tmp, gfc_conv_concat_op,
        gfc_conv_expr_op, gfc_build_compare_string,
        gfc_set_interface_mapping_bounds, gfc_conv_subref_array_arg,
        gfc_conv_derived_to_class, conv_isocbinding_procedure,
        gfc_conv_procedure_call, fill_with_spaces,
        gfc_trans_string_copy, gfc_trans_alloc_subarray_assign,
        gfc_trans_structure_assign, gfc_trans_pointer_assignment,
        gfc_trans_scalar_assign, gfc_trans_zero_assign,
        gfc_trans_array_copy, gfc_trans_array_constructor_copy): Change
        fold_build[0-9] to fold_build[0-9]_loc.
        * trans-io.c (set_parameter_const, set_parameter_value,
        set_parameter_ref, gfc_convert_array_to_string, set_string,
        set_internal_unit, io_result, set_error_locus,
        nml_get_addr_expr, build_dt): Ditto.
        * trans-openmp.c (gfc_omp_clause_default_ctor,
        gfc_omp_clause_copy_ctor, gfc_omp_clause_assign_op,
        gfc_trans_omp_array_reduction, gfc_trans_omp_atomic,
        gfc_trans_omp_do): Ditto.
        * trans.c (gfc_add_modify, gfc_build_addr_expr,
        gfc_build_array_ref, gfc_trans_runtime_error_vararg,
        gfc_trans_runtime_check, gfc_call_malloc,
        gfc_allocate_with_status, gfc_allocate_array_with_status,
        gfc_call_free, gfc_deallocate_with_status,
        gfc_call_realloc): Ditto.


Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/trans-expr.c
    trunk/gcc/fortran/trans-intrinsic.c
    trunk/gcc/fortran/trans-io.c
    trunk/gcc/fortran/trans-openmp.c
    trunk/gcc/fortran/trans.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[burnus at gcc dot gnu dot org]

------- Comment #12 from burnus at gcc dot gnu dot org  2010-09-02 14:04 -------
Subject: Bug 45186

Author: burnus
Date: Thu Sep  2 14:04:06 2010
New Revision: 163776

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=163776
Log:
2010-09-02  Tobias Burnus  <burnus@net-b.de>

        PR fortran/45186
        * trans-common.c (create_common): Change build[0-9] to
        build[0-9]_loc.
        * trans-const.c (gfc_conv_constant_to_tree,
        gfc_conv_constant_to_tree): Ditto.
        * trans-decl.c (gfc_build_qualified_array, build_entry_thunks,
        gfc_get_fake_result_decl, gfc_trans_auto_character_variable,
        add_argument_checking, create_main_function,
        gfc_generate_return): Ditto.
        * trans-types.c (gfc_get_dtype, gfc_get_array_type_bounds):
        * Ditto.
        * trans-stmt.c (allocate_temp_for_forall_nest_1,
        compute_inner_temp_size, compute_overall_iter_number,
        generate_loop_for_rhs_to_temp, generate_loop_for_temp_to_lhs,
        gfc_conv_elemental_dependencies, gfc_do_allocate,
        gfc_evaluate_where_mask, gfc_trans_allocate,
        gfc_trans_arithmetic_if, gfc_trans_call,
        gfc_trans_character_select, gfc_trans_deallocate,
        gfc_trans_do, gfc_trans_do_while, gfc_trans_forall_1,
        gfc_trans_forall_loop, gfc_trans_goto, gfc_trans_if_1,
        gfc_trans_integer_select, gfc_trans_logical_select,
        gfc_trans_pointer_assign_need_temp, gfc_trans_return,
        gfc_trans_simple_do, gfc_trans_sync, gfc_trans_where_2,
        gfc_trans_where_assign) Ditto.


Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/trans-common.c
    trunk/gcc/fortran/trans-const.c
    trunk/gcc/fortran/trans-decl.c
    trunk/gcc/fortran/trans-stmt.c
    trunk/gcc/fortran/trans-types.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[jv244 at cam dot ac dot uk]

-- 

jv244 at cam dot ac dot uk changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.5.2                       |4.6.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186

[Bug fortran/45186] [4.6 Regression] Gfortran 4.5.0 emits wrong linenumbers

[jakub at gcc dot gnu dot org]

------- Comment #11 from jakub at gcc dot gnu dot org  2010-08-17 09:25 -------
Should be fixed for 4.5.2, 4.6 will use a different approach.


-- 

jakub at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.5/4.6 Regression]        |[4.6 Regression] Gfortran
                   |Gfortran 4.5.0 emits wrong  |4.5.0 emits wrong
                   |linenumbers                 |linenumbers


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45186