Armin Rigo arigo at tunes.org
Sun Aug 8 00:23:57 CEST 2004

• Previous message: [Python-Dev] new failing test -- test_compiler
• Next message: [Python-Dev] new failing test -- test_compiler
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

---

Hi,

On Sat, Aug 07, 2004 at 09:47:22PM +0100, Michael Hudson wrote:
>         k = ''
>         for node in nodelist:
>             k += self.decode_literal(node[1])
> 
> what does this make you think of?  Yes, it's the new += string
> optimizations; Python gets upset if you mutate interned strings...
> 
> Armin is on the case...

Ahem, oups, yes, I am to blame for this memory corruption bug.  In the +=
string optimization I forgot that the string's ob_refcnt field is not always
"correct".  String interning messes with its value.  What occurred here is
that we mutated a string which had only one reference left -- that is, one
apart from the two that stringobject.c artificially removes to ob_refcnt when
interning.

This is now fixed by skipping the optimization if we got an interned string.  
While messing with ob_refcnt is always a bit fragile I am still convinced that
the += optimization is safe again.


A bientôt,

Armin.

---

• Previous message: [Python-Dev] new failing test -- test_compiler
• Next message: [Python-Dev] new failing test -- test_compiler
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

---

More information about the Python-Dev mailing list