Fix misunderstanding in Kitty kbd proto docs

doc/man/man3/notcurses_input.3.md

@@ -37,7 +37,9 @@ typedef struct ncinput {
   ncintype_e evtype;
   unsigned modifiers;// bitmask over NCKEY_MOD_*
   int ypx, xpx;      // pixel offsets within cell, -1 for undefined
-  char utf8_eff[5];  // Effective utf8 representation, taking modifier keys into account
+  uint32_t eff_text[5];  // Effective utf32 representation, taking 
+                     // modifier keys into account. This can be multiple
+                     // codepoints. Array is zero-terminated.
 } ncinput;
 
 
@@ -243,11 +245,12 @@ issues are resolved. You can determine whether the protocol is in use
 by examining the output of **notcurses-info(1)**. If the **kbd** property
 is indicated, you're using the Kitty protocol.
 
-Note that utf_eff will always return the effective text value of the key,
+Note that eff_text will always return the effective text value of the key,
 taking into account any meta keys pressed. (So shift-a will always return
 'A' in this field, regardless of if the Kitty keyboard disambiguation
 protocol is used.) This field generally only can be trusted on
-**NCTYPE_PRESS** and **NCTYPE_REPEAT** events, however.
+**NCTYPE_PRESS** and **NCTYPE_REPEAT** events, however. Also note that 
+this is a (zero-terminated) array as it can return multiple codepoints.
 
 Mouse events in the left margins will never be delivered to the
 application (as is intended), but mouse events in the bottom and right margins

include/notcurses/notcurses.h

@@ -1214,7 +1214,9 @@ typedef struct ncinput {
   ncintype_e evtype;
   unsigned modifiers;// bitmask over NCKEY_MOD_*
   int ypx, xpx;      // pixel offsets within cell, -1 for undefined
-  char utf8_eff[5];  // Effective utf8 representation, taking modifier keys into account
+  uint32_t eff_text[5];  // Effective utf32 representation, taking 
+                     // modifier keys into account. This can be multiple
+                     // codepoints. Array is zero-terminated.
 } ncinput;
 
 static inline bool

src/lib/in.c

@@ -797,7 +797,7 @@ kitty_functional(uint32_t val){
 }
 
 static void
-kitty_kbd_txt(inputctx* ictx, int val, int mods, unsigned *txt, int evtype){
+kitty_kbd_txt(inputctx* ictx, int val, int mods, uint32_t *txt, int evtype){
   assert(evtype >= 0);
   assert(mods >= 0);
   assert(val > 0);
@@ -830,23 +830,10 @@ kitty_kbd_txt(inputctx* ictx, int val, int mods, unsigned *txt, int evtype){
       tni.evtype = NCTYPE_UNKNOWN;
       break;
   }
-  // Validate the txt array. It needs to be a non-zero-length set of up to 4 bytes.
+  //note: if we don't set eff_text here, it will be set to .id later.
-  int txt_valid = 0;
+  if(txt && txt[0]!=0){
-  if(txt){
-    for (int i = 0 ; i<4 ; i++){
-      if(txt[i]==0) break;
-      if(txt[i]>255){
-        txt_valid = 0;
-        break;
-      }
-      txt_valid = 1;
-    }
-  }
-  //note: if we don't populate .utf8_eff here, it will be set to what becomes .utf8 in
-  //internal_get().
-  if(txt_valid){
     for(int i=0 ; i<4 ; i++){
-      tni.utf8_eff[i] = (char)txt[i];
+      tni.eff_text[i] = txt[i];
     }
   }
   load_ncinput(ictx, &tni);
@@ -875,7 +862,7 @@ kitty_cb(inputctx* ictx){
 
 static int 
 kitty_cb_atxtn(inputctx* ictx, int n, int with_event){
-  unsigned txt[5]={0};
+  uint32_t txt[5]={0};
   unsigned val = amata_next_numeric(&ictx->amata, "\x1b[", ';');
   unsigned ev = 0;
   unsigned mods = 0;
@@ -2709,9 +2696,7 @@ internal_get(inputctx* ictx, const struct timespec* ts, ncinput* ni){
     if(notcurses_ucs32_to_utf8(&ni->id, 1, (unsigned char*)ni->utf8, sizeof(ni->utf8)) < 0){
       ni->utf8[0] = 0;
     }
-    if(ni->utf8_eff[0] == 0){
+    if (ni->eff_text[0]==0) ni->eff_text[0]=ni->id;
-      memcpy(ni->utf8_eff, ni->utf8, sizeof(ni->utf8_eff));
-    }
   }
   if(++ictx->iread == ictx->isize){
     ictx->iread = 0;

src/lib/reader.c

@@ -401,7 +401,12 @@ bool ncreader_offer_input(ncreader* n, const ncinput* ni){
     return false;
   }
 
-  ncreader_write_egc(n, ni->utf8_eff);
+  for (int c=0; ni->eff_text[c]!=0; c++){
+    unsigned char egc[5]={0};
+    if(notcurses_ucs32_to_utf8(&ni->eff_text[c], 1, egc, 4)>=0){
+      ncreader_write_egc(n, (char*)egc);
+    }
+  }
   return true;
 }