Re: [Emacs-diffs] trunk r117169: * src/bytecode.c (FETCH) [BYTE_CODE_SAFE]: Check the bytecode wasn't

Re: [Emacs-diffs] trunk r117169: * src/bytecode.c (FETCH) [BYTE_CODE_SAFE]: Check the bytecode wasn't

[Daniel Colascione]

[-- Attachment #1: Type: text/plain, Size: 523 bytes --]

On 05/27/2014 04:48 PM, Stefan Monnier wrote:
> ------------------------------------------------------------
> revno: 117169
> revision-id: monnier@iro.umontreal.ca-20140527234835-nad9g3elkd8bep4b
> parent: eliz@gnu.org-20140527173117-et72fjq7lazj50mj
> committer: Stefan Monnier <monnier@iro.umontreal.ca>
> branch nick: trunk
> timestamp: Tue 2014-05-27 19:48:35 -0400
> message:
>   * src/bytecode.c (FETCH) [BYTE_CODE_SAFE]: Check the bytecode wasn't
>   relocated from under us.

When does that happen?


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]

Re: [Emacs-diffs] trunk r117169: * src/bytecode.c (FETCH) [BYTE_CODE_SAFE]: Check the bytecode wasn't

[Stefan Monnier]

> When does that happen?

The GC relocates strings to compact the string blocks, so it is
very common for strings to be relocated.  So it requires special code to
update pointers into the middle of strings, such as the "pc" in bytecode
activation frames.
This debug code was the one that let me finally track down the source of
a bug we had in the implementation of the "new handler bytecodes" (the
bug was in the handling of variables lives across setjmp/longjmp).


        Stefan