[OE-core] [oe-core][master][PATCH] Introduce mechanism to keep nativesdk* sstate in esdk

[Jaewon Lee]

Hi Paul,

> -----Original Message-----
> From: Paul Eggleton <paul.eggleton at linux.intel.com>
> Sent: Tuesday, September 17, 2019 3:05 PM
> To: Jaewon Lee <JAEWON at xilinx.com>
> Cc: openembedded-core at lists.openembedded.org; Alejandro Enedino
> Hernandez Samaniego <alejandr at xilinx.com>; Manjukumar Harthikote
> Matha <MANJUKUM at xilinx.com>; Bruce Ashfield <brucea at xilinx.com>
> Subject: Re: [OE-core] [oe-core][master][PATCH] Introduce mechanism to
> keep nativesdk* sstate in esdk
> 
> Hi Jaewon
> 
> Richard was waiting for me to review this - unfortunately another one that
> fell between the cracks - sorry about that.
[Jaewon] No  problem!

> 
> On Friday, 30 August 2019 5:13:39 AM NZST Jaewon Lee wrote:
> > > -----Original Message-----
> > > From: Jaewon Lee <jaewon.lee at xilinx.com>
> > > Sent: Monday, April 1, 2019 5:07 PM
> > > To: openembedded-core at lists.openembedded.org; Alejandro Enedino
> > > Hernandez Samaniego <alejandr at xilinx.com>; Manjukumar Harthikote
> > > Matha <MANJUKUM at xilinx.com>; Bruce Ashfield <brucea at xilinx.com>
> > > Cc: Jaewon Lee <JAEWON at xilinx.com>
> > > Subject: [oe-core][master][PATCH] Introduce mechanism to keep
> > > nativesdk* sstate in esdk
> > >
> > > Using SDK_INCLUDE_NATIVESDK flag to toggle inclusion of all
> > > nativesdk* sstate into esdk Currently locked-sigs.inc is generated
> > > during do_sdk_depends which doesn't pull in nativesdk packages.
> > > Generating another locked-sigs.inc in do_populate_sdk_ext and
> > > pruning it to only
> > > nativesdk* packages by using a modified version of the already
> > > existing function prune_locked_sigs and merging it with the current
> > > locked-sigs.inc Also adding SDK_INCLUDE_NATIVESDK tasklistfn to the
> > > logic surrounding setting tasklist file to not prune esdk sstate
> > > during creation
> > >
> > > Signed-off-by: Jaewon Lee <jaewon.lee at xilinx.com>
> 
> The commit message doesn't actually explain why you are adding this
> functionality. You explained it elsewhere (bug 13261) but it needs to be in
> here. I would also recommend adding a "Fixes [YOCTO #13261]" at the end so
> there's a reference back to the bug as well.
[Jaewon] Ok I will resend with commit fixed

> 
> > > @@ -414,7 +433,7 @@ python copy_buildsystem () {
> > >
> > >      sdk_include_toolchain = (d.getVar('SDK_INCLUDE_TOOLCHAIN') == '1')
> > >      sdk_ext_type = d.getVar('SDK_EXT_TYPE')
> > > -    if sdk_ext_type != 'minimal' or sdk_include_toolchain or derivative:
> > > +    if (sdk_ext_type != 'minimal' or sdk_include_toolchain or derivative)
> and not sdk_include_nativesdk:
> > >          # Create the filtered task list used to generate the sstate cache
> shipped with the SDK
> > >          tasklistfn = d.getVar('WORKDIR') + '/tasklist.txt'
> > >          create_filtered_tasklist(d, baseoutpath, tasklistfn,
> > > conf_initpath) @@ -
> 
> This logic change looks a bit odd. Are you sure this is correct?
[Jaewon] I kept the original logic and did a "and not" my flag 
Because if flag sdk_include_nativesdk is set I want tasklistfn to be none so esdksstate is not pruned during creation

Thanks,
Jaewon