fix plclip step up
This commit is contained in:
@@ -139,14 +139,13 @@ bool mmcollide(physent *d, float &hi, float &lo) // collide with a map
|
||||
for(int i = clentstats.firstclip; i < ents.length(); i++)
|
||||
{
|
||||
entity &e = ents[i];
|
||||
// if(e.type==CLIP || (e.type == PLCLIP && d->type == ENT_PLAYER))
|
||||
if (e.type==CLIP || (e.type == PLCLIP && (d->type == ENT_BOT || d->type == ENT_PLAYER || (d->type == ENT_BOUNCE && ((bounceent *)d)->plclipped)))) // don't allow bots to hack themselves into plclips - Bukz 2011/04/14
|
||||
if (e.type==CLIP || (e.type == PLCLIP && (d->type == ENT_BOT || d->type == ENT_PLAYER || (d->type == ENT_BOUNCE && ((bounceent *)d)->plclipped))))
|
||||
{
|
||||
bool hitarea = false;
|
||||
switch(e.attr7 & 3)
|
||||
{
|
||||
default: // classic unrotated clip, possibly tilted
|
||||
hitarea = fabs(e.x - d->o.x) < float(e.attr2) / ENTSCALE5 + d->radius && fabs(e.y - d->o.y) < float(e.attr3) / ENTSCALE5 + d->radius;
|
||||
hitarea = (fabs(e.x - d->o.x) - (float(e.attr2) / ENTSCALE5 + useradius) < 1e-5) && (fabs(e.y - d->o.y) - (float(e.attr3) / ENTSCALE5 + useradius) < 1e-5); // fix PLCLIP-step-up bug where e.g. 2.1<=2.1 was FALSE [sic!]
|
||||
break;
|
||||
case 3: // clip rotated 45°
|
||||
{
|
||||
|
||||
Reference in New Issue
Block a user