finally understood why the use-after-free bug happened

TODO.md

@@ -6,6 +6,7 @@ Files:-
 Playback issues in the example program that may relate to the decoder:-
 23 - Crunchy loud noise mixed with the actual audio data
 
+uncommon/02 -> 6 channel audio might be incorrect (not a decoder issue afaik)
 uncommon/04 -> We get a bits per sample mismatch
 uncommon/10 -> We error with Invalid Signature. I assume for these two I should look for the frame sync bits. But maybe my decoder should reject these. idk.
 uncommon/11 -> We error with Invalid Signature
@@ -34,8 +35,10 @@ Things:-
 [ ] Speed optimizations
 [ ] Seeking
 [ ] Fuzzing to test the stability of the code
-[ ] Check if the use-after-free bug happens on Windows, and if so, if the address sanitizer will show us the full stacktrace
+[x] Check if the use-after-free bug happens on Windows, and if so, if the address sanitizer will show us the full stacktrace
 	instead of skipping some procs.
+	This was caused by `read_data()` and the tee reader writing into the buffer and us using some memory that was freed while the buffer
+	resized. Solution is to use buffer data BEFORE calling `read_data()`
 
 [ ] Encoder (low priority for now)
 

example/main_linux.odin

@@ -54,14 +54,10 @@ main :: proc() {
 	// TODO: channel mapping is incorrect for anything with more than 2 channels.
     channels := flac_data.metadata.channels
 
-	chmap := alsa.pcm_get_chmap(handle)
-
-	fmt.println(chmap)
-
     md5_ctx: md5.Context
     md5.init(&md5_ctx)
     for {
-        frame, err := flac.read_next_frame(reader, flac_data)
+        frame, err := flac.read_next_frame(reader, flac_data, allocator)
         if err != nil {
             if err == .EOF {
                 break
@@ -83,12 +79,6 @@ main :: proc() {
 			}
         }
 
-		fmt.println("Channels:", frame.channels)
-
-		if frame.channels == 6 {
-			fmt.println("Sample:", len(frame.samples))
-		}
-
 		// TODO: resampling
 
 		// 8 bps audio

flac.odin

@@ -591,10 +591,6 @@ read_next_frame :: proc(r: ^Reader, flac: ^Flac, allocator := context.allocator)
     tee_r: io.Tee_Reader
 
     crc_buffer: Buffer
-    // NOTE: using the default heap allocator here causes a use-after-free bug on files subset/05, subset/06, and subset/25
-    // No idea what the cause of the bug is.
-    //
-    // Resolved without understanding the root issue. See below code for a note on this.
     buffer_init_allocator(&crc_buffer, 0, 128)
     defer buffer_destroy(&crc_buffer)
 
@@ -806,13 +802,6 @@ read_next_frame :: proc(r: ^Reader, flac: ^Flac, allocator := context.allocator)
     // Frame Footer
     //
     crc16_bytes := buffer_to_bytes(&crc_buffer)
-    // NOTE: The order of these two calls is important. If we're using the default heap allocator
-    // for crc_buffer and we call read_data before calculate_crc16 then we run into a memory bug.
-    // I'm still not sure the root cause of the bug but it disappears when they're in this order.
-    //
-    // In the below cases the bug doesn't happen and the order doesn't matter.
-    // When using the temp allocator or an arena allocator.
-    // When using the reader from the arguments instead of the tee reader.
     calculated_frame_crc := calculate_crc16(crc16_bytes)
     frame_crc := u16(read_data(r, u16be) or_return)
 

internal.odin

@@ -6,6 +6,7 @@ import "core:crypto/legacy/md5"
 FLAC_MAGIC :: 'f' << 24 | 'L' << 16 | 'a' << 8 | 'C'
 
 //odinfmt: disable
+@(rodata)
 crc8_lookup := [256]u8 {
     0x00, 0x07, 0x0E, 0x09, 0x1C, 0x1B, 0x12, 0x15,
     0x38, 0x3F, 0x36, 0x31, 0x24, 0x23, 0x2A, 0x2D,
@@ -41,6 +42,7 @@ crc8_lookup := [256]u8 {
     0xE6, 0xE1, 0xE8, 0xEF, 0xFA, 0xFD, 0xF4, 0xF3,
 }
 
+@(rodata)
 crc16_lookup := [256]u16 {
     0x0000, 0x8005, 0x800F, 0x000A, 0x801B, 0x001E, 0x0014, 0x8011,
     0x8033, 0x0036, 0x003C, 0x8039, 0x0028, 0x802D, 0x8027, 0x0022,