[Pkg-ime-devel] brise: change to arch:all?

[Osamu Aoki]

Hi,

On Wed, May 14, 2014 at 02:08:58PM -0400, Guo Yixuan wrote:
> Thank you, Qijian and Aron!
> 
> On Wed, May 14, 2014 at 1:49 PM, Aron Xu <aron at debian.org> wrote:
> >
> > No they aren't architecture-independent even they are installed to
> > /usr/share. 

It seems *yaml files of rime are all arch independent.

It seems package generates a binary dictionary from it and installs to
/usr/share/ ... if it is arch dependent generated data, it should stay
in arch specific path for multi-arch.  This is always problematic for
IME dictionary with arch dependent dictionary data.  libkkc is in the
same situation.

> Also, RIME itself does not build on many architectures,
> > and I believe this could be a more worthy priority.

If it is made to create arch independent dictionary, then we are in the
best situation.  The overhead of swapping byte order in C should be very
small since it happens within register and dictionary being huge, the
loading of dictionary is bound by memory bandwidth anyway.

But this is hassle ... so you may just patch with the byte order specific
things when memory alignment matters.  In my case, upstream did this
hard work.

Osamu

> > Thanks,
> > Aron
> 
> OK. But I don't think it's a good idea to install arch-dep files into
> /usr/share (maybe this is an upstream bug). 

Agreed.

> And, yes, I understand that
> FTBFS is way more important that this arch-dep issue. To clarify, are you
> referring to src:librime [1] or src:brise [2]?
> 
> [1] https://buildd.debian.org/status/package.php?p=librime
> [2] https://buildd.debian.org/status/package.php?p=brise
> 
> Cheers,
> 
> GUO Yixuan

> _______________________________________________
> Pkg-ime-devel mailing list
> Pkg-ime-devel at lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-ime-devel