On Sun, Feb 16, 2014 at 11:25:16AM +0000, Ethan Grammatikidis wrote:
> mischief reports (and i've confirmed) the following patch works. i'm
> almost scared to ask why.
> 
> diff -r 709e18f21cad sys/src/cmd/mothra/rdhtml.c
> --- a/sys/src/cmd/mothra/rdhtml.c       Sat Feb 15 17:18:58 2014 -0500
> +++ b/sys/src/cmd/mothra/rdhtml.c       Sat Feb 15 03:13:05 2014 -0800
> @@ -166,7 +166,7 @@
>  		g->hbufp=g->hbuf;
>  		g->ehbuf=g->hbuf+n;
>  	}
> -       c=*g->hbufp++&255;
> +       c=*g->hbufp++;
>  	if(c=='\n') g->lineno++;
>  	return c;
>  }
> 

i'm not sure why i said that was a good idea.
maybe this one isn't either.

here we actually read full runes from the fd.

this seems to fix visiting
https://www.cl.cam.ac.uk/~mgk25/ucs/examples/UTF-8-demo.txt
in mothra.


diff -r 709e18f21cad sys/src/cmd/mothra/rdhtml.c
--- a/sys/src/cmd/mothra/rdhtml.c	Sat Feb 15 17:18:58 2014 -0500
+++ b/sys/src/cmd/mothra/rdhtml.c	Sat Feb 15 04:10:34 2014 -0800
@@ -154,6 +154,7 @@
 	int n, c;
 	char err[1024];
 	if(g->hbufp==g->ehbuf){
+doread:
 		n=read(g->hfd, g->hbuf, NHBUF);
 		if(n<=0){
 			if(n<0){
@@ -166,7 +167,11 @@
 		g->hbufp=g->hbuf;
 		g->ehbuf=g->hbuf+n;
 	}
-	c=*g->hbufp++&255;
+	if(!fullrune(g->hbufp, g->ehbuf - g->hbufp)) {
+		goto doread;
+	}
+
+	g->hbufp += chartorune((Rune*)&c, g->hbufp);
 	if(c=='\n') g->lineno++;
 	return c;
 }