second, i think i've tracked down what was causing the problem with the
code snippet. the call to sub0 gets converted (gfc_conv_function_call)
before the generate_local_decls runs for param_test. this means that the
symbols for the dummy args in the call still say they're an array_type, so
the actuals given by sub0() are converted to array_type. i think this
means that the fixup done for by-value character dummies in
generate_local_decl must also be done when mapping actuals to formals in
gfc_conv_function_call. i've attached a patch that does this. it's been
bootstrapped and regtested on x86 and x86_64 linux with no new failures,
but the real test is whether they'll work for you. :-)
It doesn't look like it is working. I put in your patch (after removing
the previous trans-decl.c patch) Then I rebuilt and tested my program
and it still failed. Looking at the assembly language, there are still
shifts of the characters in sub0 before it calls param_test so it looks
like sub0 is still passing the arguments as arrays.
I am not sure why this is, I do hit the new code in trans-expr.c, before
calling gfc_conv_scalar_char_value I have:
<parm_decl 7d6cc2d8 my_char
type <array_type 7d76e230
type <integer_type 7d6d8150 char public unsigned string-flag QI
size <integer_cst 7d6c51b8 constant invariant 8>
unit size <integer_cst 7d6c51e0 constant invariant 1>
align 8 symtab 0 alias set -1 canonical type 7d6d8150 precision 8 min <integer_cst 7d6c5208 0> max <integer_cst 7d6c5230 255>
pointer_to_this <pointer_type 7d6e0690>>
string-flag QI size <integer_cst 7d6c51b8 8> unit size <integer_cst 7d6c51e0 1>
align 8 symtab 0 alias set -1 canonical type 7d76e230
domain <integer_type 7d76e1c0 type <integer_type 7d6d8310 int4>
SI
size <integer_cst 7d6c53e8 constant invariant 32>
unit size <integer_cst 7d6c50f0 constant invariant 4>
align 32 symtab 0 alias set -1 canonical type 7d76e1c0 precision 32 min <integer_cst 7d6c5898 1> max <integer_cst 7d6c5898 1>>>
readonly QI file x.f90 line 5 size <integer_cst 7d6c51b8 8> unit size <integer_cst 7d6c51e0 1>
align 8 context <function_decl 7d76dd20 param_test> initial <array_type 7d76e230> arg-type <array_type 7d76e230> chain <parm_decl 7d6cc3a8 my_char_2>>
And after calling gfc_conv_scalar_char_value I have:
<parm_decl 7d6cc2d8 my_char
type <integer_type 7d6d8150 char public unsigned string-flag QI
size <integer_cst 7d6c51b8 constant invariant 8>
unit size <integer_cst 7d6c51e0 constant invariant 1>
align 8 symtab 0 alias set -1 canonical type 7d6d8150 precision 8 min <integer_cst 7d6c5208 0> max <integer_cst 7d6c5230 255>
pointer_to_this <pointer_type 7d6e0690>>
readonly QI file x.f90 line 5 size <integer_cst 7d6c51b8 8> unit size <integer_cst 7d6c51e0 1>
align 8 context <function_decl 7d76dd20 param_test> initial <integer_type 7d6d8150 char> arg-type <integer_type 7d6d8150 char> chain <parm_decl 7d6cc3a8 my_char_2>>
But the arguments are still shifted before being passed from sub0 to
param_test. I don't know why.