# Tool to enhance ERROR STOP messages **URL:** \ **Category:** Uncategorized\ **Created:** [February 26, 2026, 8:54pm UTC](https://fortran-lang.discourse.group/t/tool-to-enhance-error-stop-messages/10749 "2026-02-26T20:54:23Z")\ **Posts on this page:** 4\ **Page:** 1
**Author:** ![Beliavsky](https://avatars.discourse-cdn.com/v4/letter/b/ba8739/32.png) [@Beliavsky](https://fortran-lang.discourse.group/u/Beliavsky)\ **Post date:** [February 26, 2026, 8:54pm UTC](https://fortran-lang.discourse.group/t/tool-to-enhance-error-stop-messages/10749/1 "2026-02-26T20:54:23Z")
`Error stop` unlike `stop` can be used in pure procedures and should be used to signal errors. [Pure-Fortran](https://github.com/Beliavsky/Pure-Fortran) has a script `xerror_loc.py` to enhance an `error stop` message by appending the line of the file, module, and procedure name and the values of the variables that triggered the error. For example, for the code ```fortran module m implicit none contains pure function dot(x, y) result(sum_xy) real, intent(in) :: x(:), y(:) real :: sum_xy if (size(x) /= size(y)) error stop "size(x) /= size(y)" sum_xy = sum(x*y) end function dot end module m program main use m implicit none print*,dot([2.0], [3.0, 4.0]) end program main ``` `python xerror_loc.py xerror_stop_dot.f90 --condition-values --run-both --out temp.f90` gives ```auto Applied 0 location tag rewrite(s) and 1 IF-block rewrite(s) across 1 file(s). Build (original-fortran): gfortran xerror_stop_dot.f90 -o xerror_stop_dot.original-fortran.exe ERROR STOP size(x) /= size(y) Build (transformed-fortran): gfortran Build (transformed-fortran): PASS Run (transformed-fortran): temp.transformed-fortran.exe Run (transformed-fortran): FAIL (exit 1) ERROR STOP size(x) /= size(y) [temp.f90::module m::function dot::line 7]: size(x) = 1, size(y) = 2 ``` In the error message for the transformed code, the line number of the error stop and the values of `size(x)` and `size(y)` are shown. You could print these before the error stop, but print is not allowed in pure functions. The new code around the error stop is ```fortran block character(len=1000) :: msg write(msg, "(a,a,g0,a,g0)") & & "size(x) /= size(y) [temp.f90::module m::function dot::line 7]", & & ": size(x) = ", size(x), ", & & size(y) = ", size(y) error stop trim(msg) end block ``` Using this tool you can get detailed information about why a program stopped without using a debugger or adding print statements (which may not be allowed). ---
**Author:** ![Beliavsky](https://avatars.discourse-cdn.com/v4/letter/b/ba8739/32.png) [@Beliavsky](https://fortran-lang.discourse.group/u/Beliavsky)\ **Post date:** [February 27, 2026, 12:39am UTC](https://fortran-lang.discourse.group/t/tool-to-enhance-error-stop-messages/10749/2 "2026-02-27T00:39:39Z")
An `error stop` may be triggered by multiple conditions, as in ```fortran module m implicit none contains pure subroutine mmult(a, b, c) real, intent(in) :: a(:,:) real, intent(in) :: b(:,:) real, intent(out) :: c(:,:) if (size(a,2) /= size(b,1) .or. size(c,1) /= size(a,1) .or. & size(c,2) /= size(b,2)) error stop c = matmul(a,b) end subroutine mmult end module m program main use m, only: mmult implicit none integer, parameter :: na1 = 3, na2 = 4, nb2 = 5 real :: a(na1, na2), b(na2, nb2), c(na1, nb2) call random_number(a) call random_number(b) call mmult(a, b, c) print*,c(1,1) call mmult(a, b(:, 2:), c) print*,c(1,1) end program main ``` `python xerror_loc.py xmmult.f90 --specific --out temp.f90 --run` gives an error message saying which condition caused termination, for example ```auto ERROR STOP size(c,2) /= size(b,2) [temp.f90::module m::subroutine mmult::line 8]: size(c,2) = 5, size(b,2) = 4 ``` for the code above. The `error stop` block in the transformed program is ```fortran block character(len=1000) :: msg character(len=*), parameter :: loc = & & "[temp.f90::module m::subroutine mmult::line 8]" if (size(a,2) /= size(b,1)) then write(msg, "(a,a,g0,a,g0)") "size(a,2) /= size(b,1) " // loc, & & ": size(a,2) = ", size(a,2), ", size(b,1) = ", size(b,1) else if (size(c,1) /= size(a,1)) then write(msg, "(a,a,g0,a,g0)") "size(c,1) /= size(a,1) " // loc, & & ": size(c,1) = ", size(c,1), ", size(a,1) = ", size(a,1) else write(msg, "(a,a,g0,a,g0)") "size(c,2) /= size(b,2) " // loc, & & ": size(c,2) = ", size(c,2), ", size(b,2) = ", size(b,2) end if error stop trim(msg) end block ``` Conditions are tested successively, and the first true condition is reported. If the original `error stop` test uses `any`, instead of repeated `.or.`, as in ```fortran if (any([size(a,2) /= size(b,1), size(c,1) /= size(a,1), & size(c,2) /= size(b,2)])) error stop ``` the same error message is given. ---
**Author:** ![jeff.irwin](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/jeff.irwin/32/3673_2.png) [@jeff.irwin](https://fortran-lang.discourse.group/u/jeff.irwin)\ **Post date:** [February 28, 2026, 9:08pm UTC](https://fortran-lang.discourse.group/t/tool-to-enhance-error-stop-messages/10749/3 "2026-02-28T21:08:24Z")
This is cool! Another way to do part of this is to use the compiler-defined macros ` __FILE__ ` and ` __LINE__ `, which are supported by many compilers, at least `gfortran` and Intel. I use that for example in unit tests here: [fynth/test/test.F90 at 9def56a846a9dd305acc788d0d3f0c7d31a943da · JeffIrwin/fynth · GitHub](https://github.com/JeffIrwin/fynth/blob/9def56a846a9dd305acc788d0d3f0c7d31a943da/test/test.F90#L96) ```auto #define TEST_EQ(val, expect, ntot) test_eq_str(val, expect, ntot, __FILE__ , __LINE__ ) integer function test_eq_str(val, expect, ntot, file_, line) character(len = *), intent(in) :: val, expect integer, intent(inout) :: ntot character(len = *), intent(in) :: file_ integer, intent(in) :: line ntot = ntot + 1 if (val == expect .and. len(val) == len(expect)) then test_eq_str = 0 else test_eq_str = 1 write(*,*) ERROR_STR//"str equality test failed at " & //file_//":"//to_str(line) write(*,*) " Received value: """, val, """" write(*,*) " Expected value: """, expect, """" end if end function test_eq_str ``` ---
**Author:** ![sanjana\_chetry](https://avatars.discourse-cdn.com/v4/letter/s/85e7bf/32.png) [@sanjana\_chetry](https://fortran-lang.discourse.group/u/sanjana_chetry)\ **Post date:** [March 1, 2026, 10:31am UTC](https://fortran-lang.discourse.group/t/tool-to-enhance-error-stop-messages/10749/4 "2026-03-01T10:31:17Z")
Hi everyone, I’m exploring this idea as a potential GSoC project and trying to understand how ERROR STOP messages are currently produced across compilers. I’m still getting familiar with the ecosystem, so I wanted to ask: in practice, where would richer context (like call stack or variable info) be most helpful? Any pointers would be greatly appreciated.