This patch deletes the line

	free(fontnames[fix]);

from /sys/src/cmd/acme/acme.c:/^rfget.

The strings fontnames[i] do not necessarily come  from malloc:
Initially, they are statically allocated.

Other code that sets fontnames[i] does not free first; for example

	/sys/src/cmd/acme/acme.c:/^threadmain

	/sys/src/cmd/acme/rows.c:/^rowloadfonts

The following seems to crash acme reliably (and does not with this
patch):

	Run it like /386/bin/acme
	to avoid any Load or command-line
	changes to fontnames.

	Put the text
		Font var /lib/font/bit/lucidasans/unicode.6.font
	in a window (not a tag).

	Execute the text with the middle button.

Notes:
Mon Dec 6 12:16:25 EST 2004 rsc
    why is this okay?  i think maybe the stuff in fontnames should be estrdup'ed.
    and then then free is okay.  right now for example there are calls to
    rfget that could copy name into the array and then the next line after
    the rfget call is free(name).  so estrdup seems required.
    and once estrdup is required the free is correct.
    

Thu Dec 9 13:09:33 EST 2004 swasey
    The small space leak did not bother me (or, presumably, the original author).
    As you suggest, it would be better to estrdup the initial values
    and establish the invariant that after that initialization,
    fontnames[i] are allocated with malloc.  Some "missing" frees could be
    added.  I can submit another patch.


Reference: /n/sources/patch/sorry/acme-fontnames
Date: Mon Dec 20 18:57:29 CET 2004
Reviewed-by: rsc

--- /sys/src/cmd/acme/acme.c	Mon Dec 20 18:57:29 2004
+++ /sys/src/cmd/acme/acme.c	Mon Dec 20 18:57:29 2004
@@ -798,7 +798,6 @@
 		if(reffonts[fix])
 			rfclose(reffonts[fix]);
 		reffonts[fix] = r;
-		free(fontnames[fix]);
 		fontnames[fix] = name;
 	}
 	if(setfont){