From 8b0de2aa1bce1db580828fe9fba7ab5aa2501376 Mon Sep 17 00:00:00 2001
From: nick black <dankamongmen@gmail.com>
Date: Sat, 2 Jan 2021 22:06:28 -0500
Subject: [PATCH] introduce ncwidth() to handle sextants on pre-U13 libcs #1274

---
 src/lib/egcpool.h  |  5 +++--
 src/lib/internal.h |  2 +-
 src/lib/layout.c   |  2 +-
 src/lib/ncwidth.h  | 18 ++++++++++++++++++
 4 files changed, 23 insertions(+), 4 deletions(-)
 create mode 100644 src/lib/ncwidth.h

diff --git a/src/lib/egcpool.h b/src/lib/egcpool.h
index 3706722cb..67d7711a5 100644
--- a/src/lib/egcpool.h
+++ b/src/lib/egcpool.h
@@ -11,6 +11,7 @@
 #include <string.h>
 #include <unigbrk.h>
 #include <stdbool.h>
+#include "ncwidth.h"
 #include "notcurses/notcurses.h"
 
 #ifdef __cplusplus
@@ -66,7 +67,7 @@ egcpool_grow(egcpool* pool, size_t len){
 // any NUL terminator. Neither the number of bytes nor columns is necessarily
 // equal to the number of decoded code points. Such are the ways of Unicode.
 // uc_is_grapheme_break() wants UTF-32, which is fine, because we need wchar_t
-// to use wcwidth() anyway FIXME except this doesn't work with 16-bit wchar_t!
+// to use ncwidth() anyway FIXME except this doesn't work with 16-bit wchar_t!
 static inline int
 utf8_egc_len(const char* gcluster, int* colcount){
   size_t ret = 0;
@@ -83,7 +84,7 @@ utf8_egc_len(const char* gcluster, int* colcount){
     if(prevw && uc_is_grapheme_break(prevw, wc)){
       break; // starts a new EGC, exit and do not claim
     }
-    int cols = wcwidth(wc);
+    int cols = ncwidth(wc);
     if(cols < 0){
       if(iswspace(wc)){ // newline or tab
         return ret + 1;
diff --git a/src/lib/internal.h b/src/lib/internal.h
index aefafc012..ccb413903 100644
--- a/src/lib/internal.h
+++ b/src/lib/internal.h
@@ -397,7 +397,7 @@ mbstr_find_codepoint(const char* s, char32_t cp, int* col){
     if(towlower(cp) == towlower(w)){
       return bytes;
     }
-    *col += wcwidth(w);
+    *col += ncwidth(w);
     bytes += r;
   }
   return -1;
diff --git a/src/lib/layout.c b/src/lib/layout.c
index dccd2e53d..f35be4b33 100644
--- a/src/lib/layout.c
+++ b/src/lib/layout.c
@@ -95,7 +95,7 @@ puttext_line(ncplane* n, ncalign_e align, const char* text, size_t* bytes){
       return cols;
     }
     b += consumed;
-    int width = wcwidth(w);
+    int width = ncwidth(w);
     if(width < 0){
       width = 0; // FIXME
     }
diff --git a/src/lib/ncwidth.h b/src/lib/ncwidth.h
new file mode 100644
index 000000000..42725f8a5
--- /dev/null
+++ b/src/lib/ncwidth.h
@@ -0,0 +1,18 @@
+#include <wchar.h>
+
+// wcwidth() might be missing some lengths (for instance, glibc didn't get
+// Unicode 13 support until 2.31). ncwidth() handles some characters on the
+// wcwidth() error path. it ought generally be used rather than wcwidth().
+static inline int
+ncwidth(wchar_t c){
+  int r = wcwidth(c);
+  if(r >= 0){
+    return r;
+  }
+  // Symbols for Legacy Computing were only added to glibc in 2.32 (along with
+  // the rest of Unicode 13). Handle them explicitly if wcwidth() failed.
+  if(c >= 0x1fb00 && c <= 0x1fb3b){
+    return 1;
+  }
+  return -1;
+}