[PATCH 0/2] devshell: make sure toolchains are populated in sysroot

[PATCH 0/2] devshell: make sure toolchains are populated in sysroot

[Chang Rebecca Swee Fun]

Hi all,

In the previous implementation, none of the build tools will be in
the sysroot if we clean the recipe and run devshell on it.
This is a fix on devshell.bbclass to ensure all necessary toolchains
are available in sysroot whenever we enter into devshell.
The second patch is to clean up some trailing whitespaces in
devshell.bbclass.

Thanks,
Rebecca


Chang Rebecca Swee Fun (2):
  devshell.bbclass: ensure toolchains available for do_devshell
  devshell.bbclass: clean up trailing whitespaces

 meta/classes/devshell.bbclass | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)

-- 
2.7.4

[PATCH 1/2] devshell.bbclass: ensure toolchains available for do_devshell

[Chang Rebecca Swee Fun]

Toolchains are required to be prepopulated in sysroot before
entering devshell.

[YOCTO #11464]

Signed-off-by: Chang Rebecca Swee Fun <rebecca.swee.fun.chang@intel.com>
---
 meta/classes/devshell.bbclass | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/meta/classes/devshell.bbclass b/meta/classes/devshell.bbclass
index 4de7ea6..770dcbd 100644
--- a/meta/classes/devshell.bbclass
+++ b/meta/classes/devshell.bbclass
@@ -32,7 +32,16 @@ python () {
        # can't manipulate the environment and variables here yet (see YOCTO #4795)
        d.setVarFlag("do_devshell", "manualfakeroot", "1")
        d.delVarFlag("do_devshell", "fakeroot")
-} 
+
+    # We need to ensure that the toolchains are available in the sysroot
+    # when running devshell (see YOCTO #11464)
+    # We get the dependency for do_configure depends varflag and DEPENDS
+    # variable, then append it to do_devshell depends varflag
+    deps = d.getVarFlag('do_configure', 'depends') or ""
+    for x in bb.utils.explode_deps(d.getVar('DEPENDS') or ""):
+        deps += ' %s:do_populate_sysroot' % x
+    d.appendVarFlag('do_devshell', 'depends', deps)
+}
 
 def devpyshell(d):
 
-- 
2.7.4

[PATCH 2/2] devshell.bbclass: clean up trailing whitespaces

[Chang Rebecca Swee Fun]

Signed-off-by: Chang Rebecca Swee Fun <rebecca.swee.fun.chang@intel.com>
---
 meta/classes/devshell.bbclass | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/meta/classes/devshell.bbclass b/meta/classes/devshell.bbclass
index 770dcbd..829c54c 100644
--- a/meta/classes/devshell.bbclass
+++ b/meta/classes/devshell.bbclass
@@ -8,7 +8,7 @@ python do_devshell () {
        fakeenv = d.getVar("FAKEROOTENV").split()
        for f in fakeenv:
             k = f.split("=")
-            d.setVar(k[0], k[1])           
+            d.setVar(k[0], k[1])
             d.appendVar("OE_TERMINAL_EXPORTS", " " + k[0])
        d.delVarFlag("do_devshell", "fakeroot")
 
@@ -58,7 +58,7 @@ def devpyshell(d):
         old[3] = old[3] &~ termios.ECHO &~ termios.ICANON
         # &~ termios.ISIG
         termios.tcsetattr(fd, termios.TCSADRAIN, old)
-    
+
     # No echo or buffering over the pty
     noechoicanon(s)
 
@@ -154,7 +154,7 @@ python do_devpyshell() {
     try:
         devpyshell(d)
     except SystemExit:
-        # Stop the SIGTERM above causing an error exit code    
+        # Stop the SIGTERM above causing an error exit code
         return
     finally:
         return
-- 
2.7.4

Re: [PATCH 1/2] devshell.bbclass: ensure toolchains available for do_devshell

[Richard Purdie]

On Mon, 2017-05-29 at 13:16 +0800, Chang Rebecca Swee Fun wrote:
> Toolchains are required to be prepopulated in sysroot before
> entering devshell.
> 
> [YOCTO #11464]
> 
> Signed-off-by: Chang Rebecca Swee Fun 
> ---
>  meta/classes/devshell.bbclass | 11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)

This looks awfully complicated and I'm not sure this does much more
than this would:

-addtask devshell after do_patch
+addtask devshell after do_prepare_recipe_sysroot

Can someone explain to me why we want to do the below instead as I
think the effect is the same?

I did read the bug too but that didn't help answer my question.

Yes, this does mean devshell will now trigger a lot of more
dependencies and build but I think the change below would do the same
thing too.

Cheers,

Richard

> diff --git a/meta/classes/devshell.bbclass
> b/meta/classes/devshell.bbclass
> index 4de7ea6..770dcbd 100644
> --- a/meta/classes/devshell.bbclass
> +++ b/meta/classes/devshell.bbclass
> @@ -32,7 +32,16 @@ python () {
>         # can't manipulate the environment and variables here yet
> (see YOCTO #4795)
>         d.setVarFlag("do_devshell", "manualfakeroot", "1")
>         d.delVarFlag("do_devshell", "fakeroot")
> -} 
> +
> +    # We need to ensure that the toolchains are available in the
> sysroot
> +    # when running devshell (see YOCTO #11464)
> +    # We get the dependency for do_configure depends varflag and
> DEPENDS
> +    # variable, then append it to do_devshell depends varflag
> +    deps = d.getVarFlag('do_configure', 'depends') or ""
> +    for x in bb.utils.explode_deps(d.getVar('DEPENDS') or ""):
> +        deps += ' %s:do_populate_sysroot' % x
> +    d.appendVarFlag('do_devshell', 'depends', deps)
> +}
>  
>  def devpyshell(d):
>  
> -- 
> 2.7.4
> 

Re: [PATCH 1/2] devshell.bbclass: ensure toolchains available for do_devshell

[Paul Eggleton]

Hi Richard,

On Tuesday, 30 May 2017 11:31:36 AM NZST Richard Purdie wrote:
> On Mon, 2017-05-29 at 13:16 +0800, Chang Rebecca Swee Fun wrote:
> > Toolchains are required to be prepopulated in sysroot before
> > entering devshell.
> > 
> > [YOCTO #11464]
> > 
> > Signed-off-by: Chang Rebecca Swee Fun 
> > ---
> >  meta/classes/devshell.bbclass | 11 ++++++++++-
> >  1 file changed, 10 insertions(+), 1 deletion(-)
> 
> This looks awfully complicated and I'm not sure this does much more
> than this would:
> 
> -addtask devshell after do_patch
> +addtask devshell after do_prepare_recipe_sysroot
> 
> Can someone explain to me why we want to do the below instead as I
> think the effect is the same?
> 
> I did read the bug too but that didn't help answer my question.
> 
> Yes, this does mean devshell will now trigger a lot of more
> dependencies and build but I think the change below would do the same
> thing too.

I'd suggested the method Rebecca is using in the bug, but you're right, I'm 
not sure why I didn't think of doing as you suggest instead. Presumably 
however since do_prepare_recipe_sysroot doesn't depend on do_patch we would 
still need to depend on that i.e.:

addtask devshell after do_patch do_prepare_recipe_sysroot

Right?

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: [PATCH 1/2] devshell.bbclass: ensure toolchains available for do_devshell

[Richard Purdie]

On Tue, 2017-05-30 at 12:07 +1200, Paul Eggleton wrote:
> Hi Richard,
> 
> On Tuesday, 30 May 2017 11:31:36 AM NZST Richard Purdie wrote:
> > 
> > On Mon, 2017-05-29 at 13:16 +0800, Chang Rebecca Swee Fun wrote:
> > > 
> > > Toolchains are required to be prepopulated in sysroot before
> > > entering devshell.
> > > 
> > > [YOCTO #11464]
> > > 
> > > Signed-off-by: Chang Rebecca Swee Fun 
> > > ---
> > >  meta/classes/devshell.bbclass | 11 ++++++++++-
> > >  1 file changed, 10 insertions(+), 1 deletion(-)
> > This looks awfully complicated and I'm not sure this does much more
> > than this would:
> > 
> > -addtask devshell after do_patch
> > +addtask devshell after do_prepare_recipe_sysroot
> > 
> > Can someone explain to me why we want to do the below instead as I
> > think the effect is the same?
> > 
> > I did read the bug too but that didn't help answer my question.
> > 
> > Yes, this does mean devshell will now trigger a lot of more
> > dependencies and build but I think the change below would do the
> > same
> > thing too.
> I'd suggested the method Rebecca is using in the bug, but you're
> right, I'm 
> not sure why I didn't think of doing as you suggest instead.
> Presumably 
> however since do_prepare_recipe_sysroot doesn't depend on do_patch we
> would 
> still need to depend on that i.e.:
> 
> addtask devshell after do_patch do_prepare_recipe_sysroot
> 
> Right?

Yes, you're right that we still need do_patch too.

Cheers,

Richard