emacs-pretest-bug
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: GC


From: Dave Love
Subject: Re: GC
Date: Mon, 10 May 2004 19:47:41 +0100
User-agent: Gnus/5.1002 (Gnus v5.10.2) Emacs/21.2 (gnu/linux)

Stefan Monnier <address@hidden> writes:

>> and collecting marker chains without crashing.
>
> I.e. all the weak-stuff.

If I remember correctly, marker chains aren't like weak hash tables
implementation-wise because they're not Lisp objects directly, and
should be quite straightforward.  I think I just messed up somewhere.

> Maybe I can help.  Can you describe what you know of the problem?

I thought it was likely someone else might spot the problem quickly.
Collecting the chains just crashed (probably a SEGV).  I couldn't see
where the relevant datum was getting trashed, despite looking quite
hard and having fixed similar sorts of problems.  I think there's a
macro (probably mentioned in README.GC) that flips between crashing
and letting them leak (at a surprisingly high rate if I remember
correctly).  I don't remember whether I had any more notes, but I
deleted the working directory anyway.

> No, I was referring to the trick you used to hide the marker-chain from
> Boehm's GC tracing routine (thus making it a sort of weak pointer).

Bother!

[Incidentally as far as I know, the changes Boehm suggested to gcc to
allow it to be guaranteed (?) safe for conservative collectors never
got included.]




reply via email to

[Prev in Thread] Current Thread [Next in Thread]