fix: violating strict aliasing rules in most areas in the new code.

Yes, I am aware there are plenty of violations in `conf.c`, but I'll
likely fix/rewrite those when I will use it. Since there are some other
changes I think I'll want to make.

include/endian.h

@@ -1,9 +1,45 @@
 /* Copyright (c) 2025 Quinn
  * Licensed under the MIT Licence. See LICENSE for details */
-#pragma once
 
-#if __has_include(<endian.h>)
-#include <endian.h>
+#ifndef PORTABLE_ENDIAN_H
+#define PORTABLE_ENDIAN_H 1
+
+#if defined(__GNUC__)
+
+/* test for the byteswap header */
+#if __has_include(<byteswap.h>)
+#include <byteswap.h>
+#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
+#define le16toh(x) (x)
+#define le32toh(x) (x)
+#define le64toh(x) (x)
+#define htole16(x) (x)
+#define htole32(x) (x)
+#define htole64(x) (x)
+#define be16toh(x) __bswap16(x)
+#define be32toh(x) __bswap32(x)
+#define be64toh(x) __bswap64(x)
+#define htobe16(x) __bswap16(x)
+#define htobe32(x) __bswap32(x)
+#define htobe64(x) __bswap64(x)
+#elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#define le16toh(x) __bswap16(x)
+#define le32toh(x) __bswap32(x)
+#define le64toh(x) __bswap64(x)
+#define htole16(x) __bswap16(x)
+#define htole32(x) __bswap32(x)
+#define htole64(x) __bswap64(x)
+#define be16toh(x) (x)
+#define be32toh(x) (x)
+#define be64toh(x) (x)
+#define htobe16(x) (x)
+#define htobe32(x) (x)
+#define htobe64(x) (x)
+#else
+#error machine architecture unsupported! Expected either big-endian or little-endian, make sure to use a compiler which defines __BYTE_ORDER__ (like clang or gcc)
+#endif /* byte order */
+
+/* otherwise, utilise the compiler built-ins */
 #else
 #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
 #define le16toh(x) __uint16_identity(x)
@@ -33,5 +69,12 @@
 #define htobe64(x) __uint64_identity(x)
 #else
 #error machine architecture unsupported! Expected either big-endian or little-endian, make sure to use a compiler which defines __BYTE_ORDER__ (like clang or gcc)
-#endif
-#endif
+#endif /* byte order */
+
+#endif /* has byteswap.h */
+
+#else
+#error GNU C is unavailable
+#endif /* __GNUC__ */
+
+#endif /* PORTABLE_ENDIAN_H */

makefile

@@ -38,20 +38,20 @@ endif
 ifeq ($(DEBUG),1)
 PROF = dbg
 CFLAGS  += -UNDEBUG -Og -g -Wextra -Wpedantic
-CFLAGS  += $(if $(filter 1,$(ISWIN)),,-fsanitize=address,undefined)
-LDFLAGS += $(if $(filter 1,$(ISWIN)),,-fsanitize=address,undefined)
+CFLAGS  += $(if $(filter 1,$(ISWIN)),,-fsanitize=address,undefined) -ftrapv
+LDFLAGS += $(if $(filter 1,$(ISWIN)),,-fsanitize=address,undefined) -ftrapv
 # |--profile: testing
 else ifeq ($(DEBUG),test)
 PROF = test
 CFLAGS  += -UNDEBUG -O2 -g
-CFLAGS  += $(if $(filter 1,$(ISWIN)),,-fsanitize=address)
-LDFLAGS += $(if $(filter 1,$(ISWIN)),,-fsanitize=address)
+CFLAGS  += $(if $(filter 1,$(ISWIN)),,-fsanitize=address) -ftrapv
+LDFLAGS += $(if $(filter 1,$(ISWIN)),,-fsanitize=address) -ftrapv
 else
 PROF = rel
 CFLAGS += -DNDEBUG -O2
 endif
 
-CFLAGS  += $(shell pkg-config --cflags glfw3 libarchive) -Ilib/glad/include
+CFLAGS  += $(shell pkg-config --cflags glfw3 libarchive) -Ilib/include -Ilib/glad/include
 LDFLAGS += $(shell pkg-config --libs glfw3 libarchive) -lm
 
 # get source files

src/dat/mcx.c

@@ -3,22 +3,21 @@
 #include "mcx.h"
 
 #include <assert.h>
+#include <endian.h>
 #include <stdint.h>
 #include <stdlib.h>
 #include <string.h>
 #include <time.h>
 
-#include "../util/compat/endian.h"
 #include "../util/intdef.h"
 
-#define TABLE  0x2000 // table byte size
+#define TABLE  0x800  // table (total) element count
 #define SECTOR 0x1000 // sector size
 #define CHUNKS 0x400  // amount of chunks in a file
 
 /* Moves chunks `src_s` to `src_e` (inclusive) from `src`, back onto `dst`. */
-static void mvchunks(u8 *restrict buf, u8 *src, u8 *dst, int src_s, int src_e) {
+static void mvchunks(u32 *restrict table, u8 *src, u8 *dst, int src_s, int src_e) {
 	assert(src > dst);
-	u32 *table = (u32 *)buf;
 	size_t len = src - dst; // acquire the amount of bytes that we shall move
 	assert(!(len % SECTOR));
 
@@ -34,9 +33,8 @@ static void mvchunks(u8 *restrict buf, u8 *src, u8 *dst, int src_s, int src_e) {
 /* Deletes chunk `sidx` by moving chunks up to `eidx` back over `sidx` in `buf`.
  * `rmb` is an optional additional offset that can be applied, and signifies bytes already removed.
  * Returns the bytes removed by this function. */
-static size_t delchunk(u8 *restrict buf, size_t rmb, int sidx, int eidx) {
+static size_t delchunk(u8 *restrict buf, u32 *restrict table, size_t rmb, int sidx, int eidx) {
 	// load the table data
-	u32 *table = (u32 *)buf;
 	size_t slen, bidx, blen;
 	slen = be32toh(table[sidx]) & 0xFF;          // acquire the sector length of the chunk
 	bidx = (be32toh(table[sidx]) >> 8) * SECTOR; // acquire and compute the byte offset the chunk starts at
@@ -49,20 +47,25 @@ static size_t delchunk(u8 *restrict buf, size_t rmb, int sidx, int eidx) {
 	// move the succeeding chunks over the deleted chunk
 	u8 *dst = buf + bidx - rmb;
 	u8 *src = buf + bidx + blen;
-	mvchunks(buf, src, dst, sidx, eidx - 1);
+	mvchunks(table, src, dst, sidx, eidx - 1);
 	return blen;
 }
 
-/* Just call `delchunk` with the parameters and some defaults.
+/* Call `delchunk` with the parameters and some defaults. Ensuring the table is copied correctly as well.
  * This is done instead of `delchunk` being globally linked, because
  * `delchunk` requests more specific parameters, which is confusing outside this module. */
 size_t mcx_delchunk(u8 *restrict buf, int chunk) {
-	return delchunk(buf, 0, chunk, CHUNKS);
+	u32 table[TABLE];
+	memcpy(table, buf, sizeof(table));
+	size_t res = delchunk(buf, table, 0, chunk, CHUNKS);
+	memcpy(buf, table, sizeof(table));
+	return res;
 }
 
 size_t mcx_delchunk_range(u8 *restrict buf, int start, int end) {
 	assert(start < end && end < CHUNKS);
-	u32 *table = (u32 *)buf;
+	u32 table[TABLE];
+	memcpy(table, buf, sizeof(table));
 	u8 *dst = buf + (be32toh(table[start]) >> 8) * SECTOR;
 	u8 *src = buf + (be32toh(table[end]) >> 8) * SECTOR;
 	src += (be32toh(table[end]) & 0xFF) * SECTOR;
@@ -76,7 +79,8 @@ size_t mcx_delchunk_range(u8 *restrict buf, int start, int end) {
 
 	// move the remaining chunks down
 	if (end < (CHUNKS - 1))
-		mvchunks(buf, src, dst, end, (CHUNKS - 1));
+		mvchunks(table, src, dst, end, (CHUNKS - 1));
+	memcpy(buf, table, sizeof(table));
 	return src - dst;
 }
 
@@ -96,9 +100,14 @@ size_t mcx_delchunk_bulk(u8 *restrict buf, const u16 *restrict chunks, int chunk
 	qsort(chunkids, chunkc, sizeof(int), cmp_chunkids);
 	chunkids[chunkc] = CHUNKS; // set the spare chunk to the max chunks, so the rest of the chunks are moved
 
+	u32 table[TABLE];
+	memcpy(table, buf, sizeof(table));
+
 	size_t rmb = 0;
 	for (int i = 0; i < chunkc; i++)
-		rmb += delchunk(buf, rmb, chunkids[i], chunkids[i + 1]);
+		rmb += delchunk(buf, table, rmb, chunkids[i], chunkids[i + 1]);
+
+	memcpy(buf, table, sizeof(table));
 	return rmb;
 }
 
@@ -108,5 +117,5 @@ size_t mcx_calcsize(const u8 *restrict buf) {
 	size_t size = 0;
 	for (uint i = 0; i < CHUNKS; i++)
 		size += *(buf + (i * 4) + 3);
-	return (size * CHUNKS) + TABLE;
+	return (size * CHUNKS) + (TABLE * 4);
 }

src/dat/nbt.c

@@ -3,37 +3,61 @@
 #include "nbt.h"
 
 #include <assert.h>
+#include <endian.h>
 #include <stddef.h>
 #include <stdlib.h>
 #include <string.h>
 
-#include "../util/compat/endian.h"
 #include "../util/intdef.h"
 
 #define MAX_DEPTH 512
 
+/* Extracts a big endian 16 bit integer from address `buf`, converts it to host byte size if needed and returns. */
+static inline u16 buftoh16(const void *restrict buf) {
+	u16 i;
+	memcpy(&i, buf, sizeof(i));
+	return be16toh(i);
+}
+
+/* Extracts a big endian 32 bit integer from address `buf`, converts it to host byte size if needed and returns. */
+static inline u32 buftoh32(const void *restrict buf) {
+	u32 i;
+	memcpy(&i, buf, sizeof(i));
+	return be32toh(i);
+}
+
+/* Extracts a big endian 64 bit integer from address `buf`, converts it to host byte size if needed and returns. */
+static inline u64 buftoh64(const void *restrict buf) {
+	u64 i;
+	memcpy(&i, buf, sizeof(i));
+	return be64toh(i);
+}
+
 /* Processes the incoming array data in `buf`. Which contains `nmem` items of `size`.
  * The data shall be converted to little-endian on little-endian systems
  * Outputs the allocated data to `out`, returns where the next pointer would be. */
-static const u8 *procarr(const u8 *restrict buf, i32 nmem, uint size, struct nbt_array *restrict *restrict out) {
-	size_t len = nmem * size;
-	*out = malloc(sizeof(struct nbt_array) + len);
-	if (!*out) return buf + len;
+static const u8 *procarr(const u8 *restrict buf, i32 nmemb, uint size, struct nbt_array *restrict out) {
+	size_t len = nmemb * size;
+	*out = (struct nbt_array){
+		out->nmemb = nmemb,
+		out->dat = malloc(len),
+	};
+	if (!out->dat)
+		return buf + len;
 
-	memcpy((*out)->dat, buf, len);
-	(*out)->len = nmem;
+	memcpy(out->dat, buf, len);
 	buf += len;
 
 	/* Only include this code for little-endian systems. Since only they require this logic.
 	 * Producing optimised code for other platforms. */
 #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
 	if (size == 1) return buf;
-	size_t i = 0;
-	while (i < len) {
+	i32 i = 0;
+	while (i < nmemb) {
 		switch (size) {
-		case 2:  *(u16 *)((*out)->dat + i) = be16toh(*(u16 *)((*out)->dat + i)); break;
-		case 4:  *(u32 *)((*out)->dat + i) = be32toh(*(u32 *)((*out)->dat + i)); break;
-		case 8:  *(u64 *)((*out)->dat + i) = be64toh(*(u64 *)((*out)->dat + i)); break;
+		case 2:  ((u16 *)out->dat)[i] = be16toh(((u16 *)out->dat)[i]); break;
+		case 4:  ((u32 *)out->dat)[i] = be16toh(((u32 *)out->dat)[i]); break;
+		case 8:  ((u64 *)out->dat)[i] = be16toh(((u64 *)out->dat)[i]); break;
 		default: __builtin_unreachable(); // this should be impossible
 		}
 		i += size;
@@ -43,12 +67,10 @@ static const u8 *procarr(const u8 *restrict buf, i32 nmem, uint size, struct nbt
 }
 
 /* calls `procarr` for the simple types available. */
-static const u8 *proclist(const u8 *restrict buf, struct nbt_array *restrict *restrict out) {
+static const u8 *proclist(const u8 *restrict buf, struct nbt_array *restrict out) {
 	uint size;
 
-	*out = NULL;
-
-	switch (*buf) {
+	switch (*(u8 *)buf) {
 	case NBT_I8:  size = 1; break;
 	case NBT_I16: size = 2; break;
 	case NBT_I32: // fall through
@@ -59,7 +81,9 @@ static const u8 *proclist(const u8 *restrict buf, struct nbt_array *restrict *re
 	}
 
 	buf++;
-	i32 len = (i32)be32toh(*(u32 *)buf);
+	i32 len;
+	memcpy(&len, buf, 4);
+	len = be32toh(len);
 	buf += 4;
 	return procarr(buf, len, size, out);
 }
@@ -73,25 +97,25 @@ const u8 *nbt_proctag(const u8 *restrict buf, u16 slen, void *restrict out) {
 
 	switch (*buf) {
 	case NBT_I8:  *(u8 *)out = *ptr; return ptr + 1;
-	case NBT_I16: *(u16 *)out = be16toh(*(u16 *)ptr); return ptr + 2;
+	case NBT_I16: *(u16 *)out = buftoh16(ptr); return ptr + 2;
 	case NBT_I32: // fall through
-	case NBT_F32: *(u32 *)out = be16toh(*(u32 *)ptr); return ptr + 4;
+	case NBT_F32: *(u32 *)out = buftoh32(ptr); return ptr + 4;
 	case NBT_I64: // fall through
-	case NBT_F64: *(u64 *)out = be16toh(*(u64 *)ptr); return ptr + 8;
+	case NBT_F64: *(u64 *)out = buftoh64(ptr); return ptr + 8;
 
-	case NBT_STR:     nmem = be16toh(*(u16 *)ptr), size = 1, ptr += 2; break;
-	case NBT_ARR_I8:  nmem = be32toh(*(u32 *)ptr), size = 1, ptr += 4; break;
-	case NBT_ARR_I32: nmem = be32toh(*(u32 *)ptr), size = 4, ptr += 4; break;
-	case NBT_ARR_I64: nmem = be32toh(*(u32 *)ptr), size = 8, ptr += 4; break;
+	case NBT_STR:     nmem = buftoh16(ptr), size = 1, ptr += 2; break;
+	case NBT_ARR_I8:  nmem = buftoh32(ptr), size = 1, ptr += 4; break;
+	case NBT_ARR_I32: nmem = buftoh32(ptr), size = 4, ptr += 4; break;
+	case NBT_ARR_I64: nmem = buftoh32(ptr), size = 8, ptr += 4; break;
 
 	case NBT_LIST:
-		return proclist(ptr, (struct nbt_array **)out);
+		return proclist(ptr, (struct nbt_array *)out);
 		return tmp;
 
 	default: return NULL;
 	}
 
-	return procarr(ptr, nmem, size, (struct nbt_array **)out);
+	return procarr(ptr, nmem, size, (struct nbt_array *)out);
 }
 
 
@@ -104,17 +128,17 @@ static const u8 *nexttag_list(const u8 *restrict ptr, uint *restrict const dpt,
 	ptr++;
 	switch (*tag) {
 	case NBT_END: break;
-	case NBT_I8:  ptr += (i32)be32toh(*(u32 *)ptr) * 1; break;
-	case NBT_I16: ptr += (i32)be32toh(*(u32 *)ptr) * 2; break;
+	case NBT_I8:  ptr += (i32)buftoh32(ptr) * 1; break;
+	case NBT_I16: ptr += (i32)buftoh32(ptr) * 2; break;
 	case NBT_I32: // fall through
-	case NBT_F32: ptr += (i32)be32toh(*(u32 *)ptr) * 4; break;
+	case NBT_F32: ptr += (i32)buftoh32(ptr) * 4; break;
 	case NBT_I64: // fall through
-	case NBT_F64: ptr += (i32)be32toh(*(u32 *)ptr) * 8; break;
+	case NBT_F64: ptr += (i32)buftoh32(ptr) * 8; break;
 	default:
 		// TODO: handle out of bounds... Might not be required if we use flexible array member
 		(*dpt)++;
 		tags[*dpt] = *tag;
-		lens[*dpt] = (i32)be32toh(*(u32 *)ptr);
+		lens[*dpt] = (i32)buftoh32(ptr);
 		break;
 	}
 	ptr += 4;
@@ -136,7 +160,7 @@ static const u8 *nexttag(const u8 *restrict tag, uint *restrict const dpt, i32 *
 		*dpt -= !lens[*dpt];
 	} else {
 		type = *tag;
-		ptr += be16toh(*(u16 *)(tag + 1)) + 3;
+		ptr += buftoh16(tag + 1) + 3;
 	}
 
 	switch (type) {
@@ -147,10 +171,10 @@ static const u8 *nexttag(const u8 *restrict tag, uint *restrict const dpt, i32 *
 	case NBT_I64: // fall through
 	case NBT_F64: ptr += 8; break;
 
-	case NBT_ARR_I8:  ptr += 4 + (i32)be32toh(*(u32 *)ptr) * 1; break;
-	case NBT_ARR_I32: ptr += 4 + (i32)be32toh(*(u32 *)ptr) * 4; break;
-	case NBT_ARR_I64: ptr += 4 + (i32)be32toh(*(u32 *)ptr) * 8; break;
-	case NBT_STR:     ptr += 2 + (u16)be16toh(*(u16 *)ptr) * 1; break;
+	case NBT_ARR_I8:  ptr += 4 + (i32)buftoh32(ptr) * 1; break;
+	case NBT_ARR_I32: ptr += 4 + (i32)buftoh32(ptr) * 4; break;
+	case NBT_ARR_I64: ptr += 4 + (i32)buftoh32(ptr) * 8; break;
+	case NBT_STR:     ptr += 2 + (u16)buftoh16(ptr) * 1; break;
 
 	case NBT_END:      (*dpt)--; break;
 	case NBT_COMPOUND: (*dpt)++; break;

src/dat/nbt.h

@@ -3,11 +3,11 @@
 #pragma once
 
 #include <assert.h>
+#include <endian.h>
 #include <stdbool.h>
 #include <stdlib.h>
 
 #include "../util/atrb.h"
-#include "../util/compat/endian.h"
 #include "../util/intdef.h"
 
 /* NBT (named binary tag) is a tree data structure. Tags have a numeric type ID, name and a payload.
@@ -37,8 +37,8 @@ enum nbt_tagid {
 };
 
 struct nbt_array {
-	i32 len;
-	u8 dat[];
+	i32 nmemb;
+	void *dat;
 };
 
 

src/main.c

@@ -31,13 +31,15 @@ static inline int init(void) {
 	return 0;
 }
 
-static inline void quit(void) {
+static void quit(void) {
 	glfwTerminate();
 }
 
+
 int main(int argc, char **argv) {
 	(void)argc, (void)argv;
 	printf("debug: [DBG], info: [INF], warning: [WAR], error: [ERR], fatal: [FAT]\n");
+	atexit(quit);
 	if (init()) fatal("failed to initialize!");
 
 	window_loop();

src/util/colour32.h

@@ -1,27 +0,0 @@
-/* Copyright (c) 2025 Quinn
- * Licensed under the MIT Licence. See LICENSE for details */
-#pragma once
-#include <stdint.h>
-
-#include "vec.h"
-
-#define COLOUR32_BLACK   ((u8vec4){0x00, 0x00, 0x00, 0xFF})
-#define COLOUR32_RED     ((u8vec4){0xFF, 0x00, 0x00, 0xFF})
-#define COLOUR32_YELLOW  ((u8vec4){0xFF, 0xFF, 0x00, 0xFF})
-#define COLOUR32_ORANGE  ((u8vec4){0xFF, 0x6D, 0x00, 0xFF})
-#define COLOUR32_GREEN   ((u8vec4){0x00, 0xFF, 0x00, 0xFF})
-#define COLOUR32_CYAN    ((u8vec4){0x00, 0xFF, 0xFF, 0xFF})
-#define COLOUR32_BLUE    ((u8vec4){0x00, 0x00, 0xFF, 0xFF})
-#define COLOUR32_MAGENTA ((u8vec4){0xFF, 0x00, 0xFF, 0xFF})
-#define COLOUR32_WHITE   ((u8vec4){0xFF, 0xFF, 0xFF, 0xFF})
-
-// american macros:
-#define COLOR32_BLACK   COLOUR32_BLACK
-#define COLOR32_RED     COLOUR32_RED
-#define COLOR32_YELLOW  COLOUR32_YELLOW
-#define COLOR32_ORANGE  COLOUR32_ORANGE
-#define COLOR32_GREEN   COLOUR32_GREEN
-#define COLOR32_CYAN    COLOUR32_CYAN
-#define COLOR32_BLUE    COLOUR32_BLUE
-#define COLOR32_MAGENTA COLOUR32_MAGENTA
-#define COLOR32_WHITE   COLOUR32_WHITE