From 69b5029de5a4f28b88b84d0b8dc07549c7eb3ceb Mon Sep 17 00:00:00 2001 From: Bill Currie Date: Wed, 26 Feb 2020 17:46:53 +0900 Subject: [PATCH] Throw away function parameter type alias info typedef is meant to create a simple renaming of a potentially complex type, not create a new type. Keeping the parameter type alias info makes the types effectively different when it comes to overloaded function resolution, which is quite contrary to the goal. Does expose some breakage elsewhere, though. --- tools/qfcc/TODO | 1 + tools/qfcc/source/function.c | 5 ++++- 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/tools/qfcc/TODO b/tools/qfcc/TODO index 5bce50696..8180ed95d 100644 --- a/tools/qfcc/TODO +++ b/tools/qfcc/TODO @@ -18,5 +18,6 @@ o isset() intrinsic for more consistent string handling. o arrays in entities o optional arguments for functions (alternative to overloading) vector(vector fwd, optional vector up) vectoangles = #51; +o rewrite type system to be const-correct (hard!) ? try to reduce memory consumption ?? embedded nul characters in strings (why?) diff --git a/tools/qfcc/source/function.c b/tools/qfcc/source/function.c index 538bb8f65..eb376e0ea 100644 --- a/tools/qfcc/source/function.c +++ b/tools/qfcc/source/function.c @@ -177,7 +177,10 @@ parse_params (type_t *type, param_t *parms) internal_error (0, 0); new->t.func.num_params = -(new->t.func.num_params + 1); } else if (p->type) { - new->t.func.param_types[new->t.func.num_params] = p->type; + // FIXME this cast should not be necessary, but making the + // type system const-correct would probably take a rewrite + type_t *t = (type_t *) unalias_type (p->type); + new->t.func.param_types[new->t.func.num_params] = t; new->t.func.num_params++; } }