Launcher refactoring

[Daft-Freak]

Started off trying to run the launcher on pico things. Then got lost in the launcher code so this happened.

A few of these are bug fixes (most importantly 2b84bd2). First four are entirely for squishing the launcher onto a picosystem.

Now I need to make scan_flash go away... (Shouldn't be too hard, firmware already has a list of where things are for other reasons)

[Gadgetoid]

👀

Lots of very interesting (and much needed) tweaks!

[Daft-Freak]

Now I need to make scan_flash go away... (Shouldn't be too hard, firmware already has a list of where things are for other reasons)

Huh, really not that bad:

diff --git a/32blit/engine/api_private.hpp b/32blit/engine/api_private.hpp
index fa6dd034..88599e9c 100644
--- a/32blit/engine/api_private.hpp
+++ b/32blit/engine/api_private.hpp
@@ -128,6 +128,9 @@ namespace blit {
     bool (*set_raw_cdc_enabled)(bool enabled);
     void (*cdc_write)(const uint8_t *data, uint16_t len);
     uint16_t (*cdc_read)(uint8_t *data, uint16_t len);
+
+    // another launcher API
+    void (*list_installed_games)(std::function<void(const uint8_t *, uint32_t, uint32_t)> callback);
   };
   #pragma pack(pop)
 
diff --git a/firmware/firmware.cpp b/firmware/firmware.cpp
index 2bcd9570..87af67ec 100644
--- a/firmware/firmware.cpp
+++ b/firmware/firmware.cpp
@@ -545,6 +545,11 @@ static void *get_type_handler_metadata(const char *filetype) {
   return nullptr;
 }
 
+static void list_installed_games(std::function<void(const uint8_t *, uint32_t, uint32_t)> callback) {
+  for(auto &game : game_list)
+    callback((const uint8_t *)(qspi_flash_address + game.offset), game.offset / qspi_flash_sector_size, game.size);
+}
+
 static const uint8_t *flash_to_tmp(const std::string &filename, uint32_t &size) {
   // one file at a time
   // TODO: this could be improved
@@ -630,6 +635,8 @@ void init() {
   api.flash_to_tmp = flash_to_tmp;
   api.tmp_file_closed = tmp_file_closed;
 
+  api.list_installed_games = list_installed_games;
+
   scan_flash();
   flash_scanned = true;
 
diff --git a/launcher/launcher.cpp b/launcher/launcher.cpp
index 4fb50a1b..0de7656f 100644
--- a/launcher/launcher.cpp
+++ b/launcher/launcher.cpp
@@ -342,31 +342,11 @@ void init_lists() {
 }
 
 void scan_flash() {
-#ifdef TARGET_32BLIT_HW
-  const uint32_t qspi_flash_sector_size = 64 * 1024;
-  const uint32_t qspi_flash_size = 32768 * 1024;
-  const uint32_t qspi_flash_address = 0x90000000;
-
-  for(uint32_t offset = 0; offset < qspi_flash_size;) {
-    auto header = *(BlitGameHeader *)(qspi_flash_address + offset);
-
-    if(header.magic != blit_game_magic) {
-      offset += qspi_flash_sector_size;
-      continue;
-    }
-
-    uint32_t size = header.end - qspi_flash_address;
-
-    // tiny bit of metadata parsing just to get the size
-    auto buf = (char *)(qspi_flash_address + offset + size);
-    if(memcmp(buf, "BLITMETA", 8) == 0)
-      size += *(uint16_t *)(buf + 8) + 10;
-
-    File::add_buffer_file("flash:/" + std::to_string(offset / qspi_flash_sector_size) + ".blit", (uint8_t *)(qspi_flash_address + offset), size);
-
-    offset += calc_num_blocks(size) * qspi_flash_sector_size;
+  if(api.list_installed_games) {
+    api.list_installed_games([](const uint8_t *ptr, uint32_t block, uint32_t size){
+      File::add_buffer_file("flash:/" + std::to_string(block) + ".blit", ptr, size);
+    });
   }
-#endif
 }
 
 void delete_current_game() {

(wrote it on the wrong branch though)

[Daft-Freak]

Okay device-specific code gone, but... api.erase_game takes an offset in bytes, api.launch has a special case for flash:/[offset in blocks] sigh

The new list_installed_games API returns the offset in blocks, existing code in the launcher multiplies it by the block size for erase. Should bump the API minor version here... but if we don't the worst that happens is that new launchers have an empty installed list on old firmware. Also I forgot to bump it for the previous changes, whoops.

[Daft-Freak]

I think this just needs a minor API version bump and maybe some testing.

Can try to sort out the existing API offset inconsistency later...

[Gadgetoid]

I think it's reasonable to merge this and go from there. I did a cursory firmware/launcher update and things seem to still work as expected. Thank you!!

[Daft-Freak]

... and I didn't do the version bump again, whoops.