GNU bug report logs - #58158
29.0.50; [overlay] Interval tree iteration considered harmful

Previous Next

Full log

View this message in rfc822 format

From: Gerd Möllmann <gerd.moellmann <at> gmail.com>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: 58158 <at> debbugs.gnu.org
Subject: bug#58158: 29.0.50; [overlay] Interval tree iteration considered harmful
Date: Thu, 29 Sep 2022 09:03:17 +0200

Eli Zaretskii <eliz <at> gnu.org> writes:

>> From: Gerd Möllmann <gerd.moellmann <at> gmail.com>
>> Date: Thu, 29 Sep 2022 07:29:25 +0200
>> 
>> In its current form, interval tree iteration works like this:
>> 
>> 1. Call begin_iteration(T) to iterate over tree T
>> 2. do stuff
>> 3. Call end_iteration(T)
>> 
>> with the following rules:
>> 
>> - Begin_iteration and end_iteration must be paired.
>> 
>> - There can be only one iteration per tree at a time.  Nested iteration
>>   over the same tree is not supported (abort).
>> 
>> - No GC may happen in step 2.  This is because mark_buffer iterates over
>>   buffer overlays.
>> 
>> I think this is an exceedingly dangerous design.
>
> Why, because of "no GC" requirement?  We could ensure that by calling
> inhibit_garbage_collection (if the code doesn't do that already).

It doesn't.

BTW, if anything signals in step 2, so that end_iteration isn't called,
we're also hosed.

> What higher-level operations require "interval tree iteration" that
> you describe?  Which primitives end up doing such iterations?

What has to do with overlays.  To name a few: overlay-at, overlays-in,
next-overlay-change, previous-overlay-change, overlay-lists, ...

I personally think this is a no-go.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.