Add the collation partition to the build

Includes a small linux-only bash script to generate locale partitions

CMakeLists.txt

@@ -16,10 +16,15 @@ list(APPEND EXTRA_COMPONENT_DIRS "$ENV{PROJ_PATH}/src")
 
 project(tangara)
 
+# 'collate' partition on internal flash. Contains collation data for sorting
+# strings across languages (defaults to a generic dataset based on ISO14651).
+get_filename_component(collate_full_path "tools/collate/Generic.LC_COLLATE" ABSOLUTE)
+esptool_py_flash_to_partition(flash "collate" ${collate_full_path})
+
 # /lua partition on internal flash, for storing the lua application
 spiffs_create_partition_image(lua lua FLASH_IN_PROJECT)
 
-# /repl partition on internal flash, for storing the developer repl and its deps.
+# /repl partition on internal flash, for storing the developer repl and its deps
 file(COPY lib/lua-repl/repl DESTINATION ${CMAKE_CURRENT_BINARY_DIR}/repl)
 file(COPY lib/lua-term/term DESTINATION ${CMAKE_CURRENT_BINARY_DIR}/repl)
 spiffs_create_partition_image(repl ${CMAKE_CURRENT_BINARY_DIR}/repl FLASH_IN_PROJECT)

tools/collate/README.md

@@ -0,0 +1,24 @@
+This tool, which only works on Linux, uses your system's locale data to
+generate a collator partition for Tangara. This partition is uses to sort
+database information (album names, etc.) in a way that makes sense to humans.
+
+Because this script isn't portable, `Generic.LC_COLLATE` is included as a
+pregenerated collator partition that should work well for most people's
+music libraries.
+
+## Partition format
+
+The collator partition has 3MiB reserved for it in the partition table, which
+is enough to hold the compiled collation data for ISO14651 character sorting,
+plus a small amount of free space.
+
+The first 8 bytes of the partition are reserved for a partition name; usually
+the name of the locale that the collation data was compiled for, or 'Generic'
+for ISO14651 data.
+
+Following the 8 byte name, is a complete LC_COLLATE file, as generate by GNU
+libc's `localedef` utility. Debian's version 2.37 of this utility is known to
+work.
+
+FIXME: We should vendor this version of the tool, so that our format remains
+stable across glibc changes.

tools/collate/mkcollator.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+set -eu
+
+if [ -z "${2:-}" ]; then
+  name="Generic"
+  locale="iso14651_t1"
+else
+  name="$1"
+  locale="$2"
+fi
+
+working_dir=$(mktemp -d)
+
+echo -n "$name" | dd iflag=fullblock bs=8 count=1 conv=sync of="$working_dir/name"
+localedef -f UTF-8 -i "$locale" --no-archive "$working_dir" || true
+cat "$working_dir/name" "$working_dir/LC_COLLATE" > "$name.LC_COLLATE"