Re: 51w6 upvalue GC bug

[Rici Lake <lua@...>]

On 3-Aug-05, at 9:17 AM, Rici Lake wrote:

> That's a little disturbing, because it seems to imply that something 
> on the stack might not get marked (in this case, the closure in the 
> local 'capture'). I'm sure there are subtleties that I'm not seeing, 
> though.

OK, I found the thing I was missing -- the grayagain list. So all 
stacks are re-marked in atomic(), which should take care of any mutated 
objects.

Consequently, the only problem is the loss of the g->gray list after 
upvals are re-marked.

I'm still not completely convinced about weak tables; I can't shake the 
feeling that they ought to be done after the re-marking of the stacks 
rather than before. But don't take that as gospel.

Here's the corrected patch (really, just a delete of the second 
patch-chunk; thanks, Mike). (By the way, I, for one, would appreciate a 
complete work6 patchset -- but if you don't want to post that to the 
list, I'd settle for email :)

R.

---- cut here

--- lua-5.1-work6/src/lgc.c	2005-05-05 17:34:03 +0200
+++ src/lgc.c	2005-08-03 15:17:17 +0200
@@ -528,10 +528,10 @@
 static void atomic (lua_State *L) {
   global_State *g = G(L);
   size_t udsize;  /* total size of userdata to be finalized */
-  /* remark objects cautch by write barrier */
-  propagateall(g);
   /* remark occasional upvalues of (maybe) dead threads */
   remarkupvals(g);
+  /* remark objects caught by write barrier */
+  propagateall(g);
   /* remark weak tables */
   g->gray = g->weak;
   g->weak = NULL;