Rockbox mail archiveSubject: Re: Please help test gerrit#890
Re: Please help test gerrit#890
From: Thomas Martitz <kugel_at_rockbox.org>
Date: Sun, 06 Jul 2014 21:18:46 +0200
Am 05.07.2014 10:30, schrieb Jonathan Gordon:
> On 4 July 2014 16:21, Thomas Martitz <kugel_at_rockbox.org
> <mailto:kugel_at_rockbox.org>> wrote:
> Why do you except this to fix skin issues? It only _adds_ fragile
> pointer management. What's wrong with the current aproach?
> What's wrong with the current approach is that it is extremely
> difficult to track down bugs when they happen, and that the code
> itself is quite bloated (seriously +5k to do offset after adding all
> the code to move the pointers!). This way when there is an issue we
> get a lovely data abort with a PC.
What makes you believe that you get a data abort on a PC when you
wouldn't get one now? This is a false assumption.
What you are proposing to add is the exact opposite what we did a few
years back, when we all happily agreed that the offset-approach is
better for the skin engine. Please clear me up on why you want to
reverse that choice.
And better debuggability is not true. You don't get a data abort for
"forgetting to update pointer", which is a scenario is is luckily
impossible in the current code.
Received on 2014-07-06