RFC: Should do_runstrip be removed for kernel packages?

RFC: Should do_runstrip be removed for kernel packages?

[Scott Garman]

Hello,

I've noticed for many kernel builds in OE/Poky that there are non-fatal 
runstrip errors being generated during the do_pacakge stage of the build.

This is happening because a handful of utility binaries (in arch/boot/ 
and scripts/) are being compiled for the host architecture, and then 
strip for the target architecture is being run. When these architectures 
don't match, runstrip complains and generates these errors.

Currently in Poky we are maintaining a blacklist of these binaries and 
deleting them in do_install:

http://git.pokylinux.org/cgit.cgi/poky/commit/?id=698dab01098816e4894beb4b3dad556f045c074d

This is a hack that I'd like to keep as temporary as possible.

I'm curious though if it is even common to perform do_runstrip for 
kernel packages? What are the use cases that require this? If there are 
none, then I'm inclined to simply disable do_runstrip in kernel.bbclass.

Another option - less elegant, but better than the current hack - might 
be to use a variable to disable runstrip and set that within the kernel 
recipe files.

Thoughts or other ideas?

Scott

-- 
Scott Garman
sgarman at zenlinux dot com

Re: RFC: Should do_runstrip be removed for kernel packages?

[Tom Rini]

Scott Garman wrote:
> Hello,
> 
> I've noticed for many kernel builds in OE/Poky that there are non-fatal 
> runstrip errors being generated during the do_pacakge stage of the build.
> 
> This is happening because a handful of utility binaries (in arch/boot/ 
> and scripts/) are being compiled for the host architecture, and then 
> strip for the target architecture is being run. When these architectures 
> don't match, runstrip complains and generates these errors.
> 
> Currently in Poky we are maintaining a blacklist of these binaries and 
> deleting them in do_install:
> 
> http://git.pokylinux.org/cgit.cgi/poky/commit/?id=698dab01098816e4894beb4b3dad556f045c074d 
> 
> 
> This is a hack that I'd like to keep as temporary as possible.
> 
> I'm curious though if it is even common to perform do_runstrip for 
> kernel packages? What are the use cases that require this? If there are 
> none, then I'm inclined to simply disable do_runstrip in kernel.bbclass.
> 
> Another option - less elegant, but better than the current hack - might 
> be to use a variable to disable runstrip and set that within the kernel 
> recipe files.

There's already a SKIP for strip, iirc.  Throwing that into 
kernel.bbclass or the modules bbclass would get the discussion moving 
more, perhaps (I'm a bit torn here).

-- 
Tom Rini
Mentor Graphics Corporation