Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[DragonDePlatino]

For my roguelike, I've finally started working on the game logic. Unfortunately, also I've run into an issue. To add entities like monsters and tiles to my map, I decided I'd pass an EntityID struct into each entity's constructor so I can define its default data.

entity_defs.h:

Code: [Select]

#ifndef ENTITYDEFS_H
#define ENTITYDEFS_H

#include <string>

enum solidity_t {NONSOLID, SEMISOLID, SOLID, IMPASSABLE};
enum ai_t {NULL_AI, PLAYER_AI};
enum anim_t {NULL_ANIM, HUMANOID_ANIM};

struct EntityID
{
/* Gameplay variables */
std::string name; // Plain text name like "Rogue"
enum solidity_t solidity; // What can pass through entity
enum ai_t ai; // How entity reacts to input

/* Graphical variables */
std::string image; // Name of SDL_Texture loaded in TEXTURES
int tilepos[2]; // Coordinates of default sprite on image, in pixels
enum anim_t animtype; // How entity is animated between turns
};

#endif // ENTITYID_H
entity_defs.cpp:

Code: [Select]

#include "entity_defs.h"

EntityID IDList[2]{{"NULL", NONSOLID, NULL_AI, "tiles", {0, 0}, NULL_ANIM},
{"Stone brick", SOLID, NULL_AI, "tiles", {1, 0}, NULL_ANIM}};               
Unfortunately, whenever I close my program I get a Windows APPCRASH and debugging tells me its a "segmentation fault". I'm not even including entity_defs.h anywhere or using any of the data I made here. If I comment out the IDList it fixes the problem and the program closes as normal. What am I doing wrong?

[i2amroy]

Does debugging tell you exactly where the segmentation fault is occurring? A stack trace would probably be relatively useful.

[DragonDePlatino]

Apparently, my debugger is pointing to the IDList and giving me this error:

Spoiler (click to show/hide)

I...have no idea what any of this means. I didn't know I'd need to learn ASM just to make my silly game.

[Rose]

Try Visual Studio? It's got a much better debugger.

[i2amroy]

Also it looks like you might be compiling with one of the -Ox flags, which is probably why that debugger includes things like "optimized out". Try disabling all compilation optimization flags so that you can see the actual debug information. (Lots of times if you're in an IDE this is as simple as switching the compilation target from "release" to "debug" or something similar).

[DragonDePlatino]

You know what? Screw it. I don't even know why I'm messing with arrays when I could use vectors, which I think manage their own memory. This works:

Code: [Select]

#include "entity_defs.h"

#include <vector>

std::vector<EntityID> IDList
{
{"NULL", NONSOLID, NULL_AI, "tiles", {0, 0}, NULL_ANIM},
{"Brick wall", SOLID, NULL_AI, "tiles", {32, 0}, NULL_ANIM},
{"Brick background", NONSOLID, NULL_AI, "tiles", {64, 0}, NULL_ANIM},
{"Wooden platform",     SEMISOLID,  NULL_AI,    "tiles",    {96, 0},    NULL_ANIM},
{"Wooden ladder",       SEMISOLID,  NULL_AI,    "tiles",    {128,0},    NULL_ANIM},
{"Rogue",               SOLID,      PLAYER_AI,  "monsters", {0, 0}, PLAYER_ANIM}
};(Wow, the forum really butchered my indentation there)

@i2amroy
Nope...Only flag that's set was -g3. I'll keep the release/debug thing in mind, though. I can immediately see that in my Compiler options and it could come in handy.

@Japa
If I ever get an issue like this again and I can't solve it, I'll give Visual Studio a shot. When I was learning C++ I struggled to compile in anything other than Orwell Dev-C++ (fantastic IDE for beginners) but I guess I'll need to graduate to Visual Studio soon.

[i2amroy]

but I guess I'll need to graduate to Visual Studio soon.

CLion is another pretty good C++/C IDE if you want something a little slimmer (and is doubly useful if you've used other JetBrains IDE's since it has virtually the same user interface). It does require a license though (which you can get free if you are a student or under a few other circumstances), no free version yet sadly.

[Reelya]

Well, you might want to screw with arrays if your code needs to be at all efficient with resources.

A raw array is just a bunch of data packed in memory, by itself it cannot cause a seg fault. But the stl strings used inside that array, well they are code objects, so when they are created and deallocated, actual code is called. stl strings are actually a real memory hog: just an empty stl string costs 28 bytes, before you even put anything in it.

If you have constant strings, it's a lot more efficient (in both speed and memory) to use const char *rather than std::string, and you avoid the program running actual code to allocate and deallocate them, which probably had something to do with the seg fault.

[monkey]

Look at the backtrace, it is faulting in EntityId's destructor. Methinks you were storing a reference somewhere and it got delete'd twice.
You could use Valgrind to tell you if it's a double delete, though I don't think you can locate the first [rogue] delete with it.

[Reelya]

Monkey is right about that. I'd suggest forget the vector, leave it as a raw array and fix the issue inside the EntityId class itself. Are you using any dynamic memory e.g. "new" inside the EntityId? If so you need to make sure all bases are checked: either call the new in the constructor or init function, make sure the default pointer value is set to 0 in the constructor if it's not allocated there, and for every pointer, check if it's not-null in the destructor, and call delete if it's not-null. Not doing this right accounts for by far the bulk of "crash on exits".

Don't mess with vectors for constant data, they really don't serve any purpose at all then. The only viable reason would be bounds checking, and std::vector doesn't do that for you.

[Skyrunner]

I'm not sure whether this SQL query is a bad idea.

SELECT * FROM reservations WHERE
     (startdate < '5' AND '5' < enddate)
  OR (startdate < '6' AND '6' < enddate)
  OR ('5' < startdate AND enddate < '6')
  OR (startdate < '5' AND '6' < enddate)
  OR (startdate == '5' AND enddate == '6');

The idea is to compare with given start and end times ('5' and '6' in this test-case, I was using SQL fiddle) against the database and return all that overlap. I have five groups of comparisons because there are five cases where two events (e1, e2) could overlap: e1 is partially overlapping either in front or behind e2; e1 entirely encompasses e2 or vice versa, and e1 is identical to e2.

[EnigmaticHat]

Welp, I've graduated from lists-within-lists to a dictionary-of-dictionaries.

[MagmaMcFry]

I'm not sure whether this SQL query is a bad idea.

SELECT * FROM reservations WHERE
     (startdate < '5' AND '5' < enddate)
  OR (startdate < '6' AND '6' < enddate)
  OR ('5' < startdate AND enddate < '6')
  OR (startdate < '5' AND '6' < enddate)
  OR (startdate == '5' AND enddate == '6');

The idea is to compare with given start and end times ('5' and '6' in this test-case, I was using SQL fiddle) against the database and return all that overlap. I have five groups of comparisons because there are five cases where two events (e1, e2) could overlap: e1 is partially overlapping either in front or behind e2; e1 entirely encompasses e2 or vice versa, and e1 is identical to e2.

To test whether two intervals [a,b] and [c, d] overlap, you need only two comparisons: (a < d) AND (c < b).

[Rose]

Welp, I've graduated from lists-within-lists to a dictionary-of-dictionaries.

I have dictionaries nested three deep as well as lookup tables for them.

[Orange Wizard]

What do you even do with that many dimensions