From: Thomas Martitz <kugel_at_rockbox.org>
Date: Sat, 29 Nov 2014 11:45:15 +0100

Am 29.11.2014 um 09:55 schrieb Andrey Ryabinin:
> [Duplicating [3] in mailing list for wider audience and discussion.]
>
> Commit 7d1a47cf ("Rewrite filesystem code (WIP)") introduced
> regressions ([1],[2]) on several rk27xx targets, making them nearly unusable.
>
> Due to size of this change it's nearly impossible to find the root
> of the problem, so I'm suggesting to revert it for now.
>
> I'll keep searching the source of problem. It definitely will take a
> lot of time. But once I'll find it we could revert the revert back.
>
> We don't have stable release for this targets, cause they are never
> been officially stable. So dev builds is the only option for users and
> I would like to not keep them broken yet another hell knows how many
> month.
>
> [1] http://forums.rockbox.org/index.php/topic,48658.0.html
> [2] http://forums.rockbox.org/index.php/topic,48675.msg230556.html#msg230556
> [3] http://gerrit.rockbox.org/r/#/c/1056/
>
>


I tend to agree. I think it also broke the Samsung ypr0/yrp1 targets
(though the fix for those might be trivial). But the commit message
doesn't read like it would give anything essential/substantial enough to
warrant breaking a whole class of targets.

That said, my guess is that the root cause for the rk27xx is probably
somewhere in the interaction with the hardware dependent code. It would
be nice if that was found and the filesystem commit could stay (or be
re-applied) because it does bring some benefits.

But yea, revert if it breaks a lot of targets that otherwise have no
working build.

Best regards.
Received on 2014-11-29