[#64703] Add `Hash#fetch_at` (issue #10017) — Wojtek Mach <wojtek@...>
Hey guys
1 message
2014/09/01
[#64711] [ruby-trunk - Bug #10193] [Closed] TestIO#test_readpartial_locktmp fails randomly — nobu@...
Issue #10193 has been updated by Nobuyoshi Nakada.
3 messages
2014/09/02
[#64744] [ruby-trunk - Bug #10202] [Open] TestBenchmark#test_realtime_output breaks on ARM — v.ondruch@...
Issue #10202 has been reported by Vit Ondruch.
3 messages
2014/09/03
[#64823] documenting constants — Xavier Noria <fxn@...>
I am writing a Rails guide about constant autoloading in Ruby on
5 messages
2014/09/07
[#64838] [ruby-trunk - Bug #10212] [Open] MRI is not for lambda calculus — ko1@...
Issue #10212 has been reported by Koichi Sasada.
6 messages
2014/09/08
[#64858] Re: [ruby-trunk - Bug #10212] [Open] MRI is not for lambda calculus
— Eric Wong <normalperson@...>
2014/09/08
rb_env_t may use a flexible array, helps a little even on my busy system:
[#64871] Re: [ruby-trunk - Bug #10212] [Open] MRI is not for lambda calculus
— SASADA Koichi <ko1@...>
2014/09/08
(2014/09/08 19:48), Eric Wong wrote:
[#64972] [ruby-trunk - Bug #10231] [Open] Process.detach(pid) defines new singleton classes every call — headius@...
Issue #10231 has been reported by Charles Nutter.
3 messages
2014/09/11
[#64980] [ruby-trunk - Bug #10212] MRI is not for lambda calculus — ko1@...
Issue #10212 has been updated by Koichi Sasada.
4 messages
2014/09/12
[#65142] [ruby-trunk - Feature #10267] [Open] Number of processors — akr@...
Issue #10267 has been reported by Akira Tanaka.
4 messages
2014/09/20
[#65144] Re: [ruby-trunk - Feature #10267] [Open] Number of processors
— Eric Wong <normalperson@...>
2014/09/20
akr@fsij.org wrote:
[#65210] [ruby-trunk - misc #10278] [Assigned] [RFC] st.c: use ccan linked list — nobu@...
Issue #10278 has been updated by Nobuyoshi Nakada.
3 messages
2014/09/22
[ruby-core:64847] Re: [ruby-trunk - Bug #10206] garbage symbols crash symbol GC
From:
Eric Wong <normalperson@...>
Date:
2014-09-08 08:37:50 UTC
List:
ruby-core #64847
ko1@atdot.net wrote:
> At first, Symbol is VALUE and it should be marked.
>
> So that the following code should not be allowed.
>
> ```
> id = SYM2ID(garbage_sym);
> ```
>
> In this case, afeter sweeping, garbage_sym becomes freed VALUE.
>
> What happen on it?
Looking at this more, we may run dsymbol_check too late in
dsymbol_pindown. I think we must run dsymbol_check immediately after
looking up dynamic syms from global_symbol.str_id, and not later.
I think this may be a fix (still testing):
--- a/symbol.c
+++ b/symbol.c
@@ -458,7 +458,10 @@ dsymbol_pindown(VALUE sym)
if (UNLIKELY(SYMBOL_PINNED_P(sym) == 0)) {
VALUE fstr = RSYMBOL(sym)->fstr;
- sym = dsymbol_check(sym);
+
+ if (UNLIKELY(rb_objspace_garbage_object_p(sym))) {
+ rb_bug("attempted to pindown garbage sym");
+ }
FL_SET(sym, SYMBOL_PINNED);
/* make it permanent object */
@@ -525,6 +528,9 @@ rb_intern_cstr_without_pindown(const char *name, long len, rb_encoding *enc)
OBJ_FREEZE(str);
if (st_lookup(global_symbols.str_id, str, &id)) {
+ if (ID_DYNAMIC_SYM_P((ID)id)) {
+ return (ID)dsymbol_check((VALUE)id);
+ }
return (ID)id;
}