Re: [kbd] [Fwd: Re: [Pkg-kbd-devel] kbd -b needlessly requires VT access]

On Mon, Aug 29, 2016 at 08:17:29AM -0700, Andy Valencia wrote:
> At the request of Andreas, I am forwarding this bug I reported on Debian.
> Briefly, loadkeys in -b mode is just a file utility, converting text
> format to binary.  And yet the code is structured so that a console
> FD must be opened, needlessly requiring the file conversion to be run
> in a console environment.

Thanks for report! Good catch!

http://git.kernel.org/cgit/linux/kernel/git/legion/kbd.git/commit/?id=2740f65352e3806dac56e16e5856eac5702caf6e

> ---------------------------- Original Message ----------------------------
> Subject: Re: [Pkg-kbd-devel] kbd -b needlessly requires VT access
> From:    "Andreas Henriksson" <[email protected]>
> Date:    Sun, August 28, 2016 4:59 am
> --------------------------------------------------------------------------
> 
> Hello Andy Valencia.
> 
> On Fri, Aug 26, 2016 at 08:54:53AM -0700, Andy Valencia wrote:
> > When generating a binary map using "loadkeys -b" to target a smaller
> > device, the command fails unless run in an environment where it can
> > access a console device for itself.  This is due to its unconditional
> > use of getfd(), even when all the command is going to do is write
> > out a binary version of its input file.
> 
> I see that the fd returned from getfd is only conditionally used, but
> the call to getfd will result in exit(1) being called if getfd is
> about to fail. This indeed sounds suboptimal.
> Not sure if moving the getfd call to where fd is used, or if guarding
> the early call to getfd with the same guards that will make use
> of getfd, is the best option.
> 
> Could you please try to contact the upstream development list about
> this issue?
> 
> https://lists.altlinux.org/mailman/listinfo/kbd
> 
> Regards,
> Andreas Henriksson
> 
> 
> _______________________________________________
> kbd mailing list
> [email protected]
> https://lists.altlinux.org/mailman/listinfo/kbd

-- 
Rgrds, legion