Re: [PATCH V2 2/2] fortran: Testsuite, fix different type naming across compilers.

["Heckel, Bernhard" <bernhard.heckel@intel.com>]

On 15/04/2016 16:41, Yao Qi wrote:
> Bernhard Heckel <bernhard.heckel@intel.com> writes:
>
>> -# Depending on the compiler version being used, the name of the 4-byte integer
>> -# and real types can be printed differently.  For instance, gfortran-4.1 uses
>> -# "int4" whereas gfortran-4.3 uses "int(kind=4)".
>> -set int4 "(int4|integer\\(kind=4\\))"
>> -set real4 "(real4|real\\(kind=4\\))"
>> -
>> -gdb_test "ptype p" "type = Type bar\r\n *${int4} :: c\r\n *${real4} :: d\r\n *End Type bar"
>> +gdb_test "ptype p" "type = Type bar\r\n *$int :: c\r\n *$real :: d\r\n *End Type bar"
>
>> +
>> +proc fortran_int4 {} {
>> +    if {[test_compiler_info {gcc-4-[012]-*}]} {
>> +	return "int4"
>> +    } elseif {[test_compiler_info {gcc-*}]} {
>> +	return "integer\\(kind=4\\)"
>> +    } elseif {[test_compiler_info {icc-*}]} {
>> +	return "INTEGER\\(4\\)"
>> +    } else {
>> +	return "unknown"
>> +    }
>> +}
>> +
>> +
>> +if ![info exists int4] then {
>> +    set int4 [fortran_int4]
>> +}
> Why do you prefer to define these info?  In each test, we can do
>
> set int4 [fortran_int4],
>
> IMO, that is much cleaner than your current approach (define int4
> globally in lib/fortran.exp and use it everywhere).
>
I wanted to avoid copy in the same code in several testcases but when I 
look again, it's
better to get rid of the globals here.
Intel Deutschland GmbH
Registered Address: Am Campeon 10-12, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de
Managing Directors: Christin Eisenschmid, Christian Lamprechter
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928