[PATCH] testimage: regenerate rootfs before testing

[PATCH] testimage: regenerate rootfs before testing

[Cristian Iorga]

While starting a test with testimage, changes
of the image under test don't trigger a rebuild,
so fix that.

Signed-off-by: Cristian Iorga <cristian.iorga@intel.com>
---
 meta/classes/testimage.bbclass | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta/classes/testimage.bbclass b/meta/classes/testimage.bbclass
index dc163c7..ce7f5c1 100644
--- a/meta/classes/testimage.bbclass
+++ b/meta/classes/testimage.bbclass
@@ -53,7 +53,7 @@ TESTIMAGELOCK_qemuall = ""
 python do_testimage() {
     testimage_main(d)
 }
-addtask testimage
+addtask testimage after do_rootfs
 do_testimage[nostamp] = "1"
 do_testimage[depends] += "${TESTIMAGEDEPENDS}"
 do_testimage[lockfiles] += "${TESTIMAGELOCK}"
-- 
2.1.0

Re: [PATCH] testimage: regenerate rootfs before testing

[Paul Eggleton]

On Thursday 23 April 2015 09:39:08 Cristian Iorga wrote:
> While starting a test with testimage, changes
> of the image under test don't trigger a rebuild,
> so fix that.
> 
> Signed-off-by: Cristian Iorga <cristian.iorga@intel.com>
> ---
>  meta/classes/testimage.bbclass | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/meta/classes/testimage.bbclass b/meta/classes/testimage.bbclass
> index dc163c7..ce7f5c1 100644
> --- a/meta/classes/testimage.bbclass
> +++ b/meta/classes/testimage.bbclass
> @@ -53,7 +53,7 @@ TESTIMAGELOCK_qemuall = ""
>  python do_testimage() {
>      testimage_main(d)
>  }
> -addtask testimage
> +addtask testimage after do_rootfs
>  do_testimage[nostamp] = "1"
>  do_testimage[depends] += "${TESTIMAGEDEPENDS}"
>  do_testimage[lockfiles] += "${TESTIMAGELOCK}"

I can't remember for certain, but I suspect the lack of this dependency was 
intentional, so that you could e.g. copy in an image built elsewhere and run 
tests on it. Maybe we can take the decision we don't need that, but it should 
be a conscious decision.

On the other hand, testimage-auto.bbclass does set up this dependency, 
although it also adds a dependency from do_build on do_testimage as well. 
There is a bug open to integrate testimage-auto.bbclass with testimage.bbclass 
as well (https://bugzilla.yoctoproject.org/show_bug.cgi?id=6155).

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: [PATCH] testimage: regenerate rootfs before testing

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 693 bytes --]

On 23 April 2015 at 12:20, Paul Eggleton <paul.eggleton@linux.intel.com>
wrote:

> I can't remember for certain, but I suspect the lack of this dependency was
> intentional, so that you could e.g. copy in an image built elsewhere and
> run
> tests on it. Maybe we can take the decision we don't need that, but it
> should
> be a conscious decision.
>

My guess was that the missing dependency was from when rootfs always
regenerated an image, so if you were iterating the test suite it would
always rebuild the image.  This isn't the case now.

Someone should speak with QA and see if they actually use the ability to
copy in images.  Cristian, I'm looking at you. :)

Ross

[-- Attachment #2: Type: text/html, Size: 1180 bytes --]

Re: [PATCH] testimage: regenerate rootfs before testing

[Paul Eggleton]

On Thursday 23 April 2015 16:02:45 Burton, Ross wrote:
> On 23 April 2015 at 12:20, Paul Eggleton <paul.eggleton@linux.intel.com>
> wrote:
> > I can't remember for certain, but I suspect the lack of this dependency
> > was intentional, so that you could e.g. copy in an image built elsewhere
> > and run tests on it. Maybe we can take the decision we don't need that,
> > but it should be a conscious decision.
> 
> My guess was that the missing dependency was from when rootfs always
> regenerated an image, so if you were iterating the test suite it would
> always rebuild the image.  This isn't the case now.

I don't believe so since that change was made prior to testimage's existence 
(although it was only a month or so before).

> Someone should speak with QA and see if they actually use the ability to
> copy in images.  Cristian, I'm looking at you. :)

I'm assuming our QA team aren't the only people that use testimage ;)

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: [PATCH] testimage: regenerate rootfs before testing

[Richard Purdie]

On Thu, 2015-04-23 at 14:25 +0100, Paul Eggleton wrote:
> On Thursday 23 April 2015 16:02:45 Burton, Ross wrote:
> > On 23 April 2015 at 12:20, Paul Eggleton <paul.eggleton@linux.intel.com>
> > wrote:
> > > I can't remember for certain, but I suspect the lack of this dependency
> > > was intentional, so that you could e.g. copy in an image built elsewhere
> > > and run tests on it. Maybe we can take the decision we don't need that,
> > > but it should be a conscious decision.
> > 
> > My guess was that the missing dependency was from when rootfs always
> > regenerated an image, so if you were iterating the test suite it would
> > always rebuild the image.  This isn't the case now.
> 
> I don't believe so since that change was made prior to testimage's existence 
> (although it was only a month or so before).
> 
> > Someone should speak with QA and see if they actually use the ability to
> > copy in images.  Cristian, I'm looking at you. :)
> 
> I'm assuming our QA team aren't the only people that use testimage ;)

This was discussed long ago and there are two different tasks as a
compromise. testimage_auto runs automatically after a rootfs is built
and has the correct dependencies. "testimage" is triggered manually. The
autobuilder actually uses testimage so it can be explicit about which
images get tested.

Cheers,

Richard

Re: [PATCH] testimage: regenerate rootfs before testing

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 910 bytes --]

On 23 April 2015 at 19:47, Richard Purdie <
richard.purdie@linuxfoundation.org> wrote:

> This was discussed long ago and there are two different tasks as a
> compromise. testimage_auto runs automatically after a rootfs is built
> and has the correct dependencies. "testimage" is triggered manually. The
> autobuilder actually uses testimage so it can be explicit about which
> images get tested.
>

The use-case was Cristi editing a recipe and running bitbake myimage -c
testimage with the expectation that the image gets rebuilt before testing.
This isn't unreasonable at currently the only way to do this is bitbake
myimage && bitbake myimage -c testimage.

If testimage_auto is merged with testimage I presume there'll be some
variable to control the behaviour.  Could we extend it to have an option to
make the image depend on the rootfs generation to cover all three use-cases?

Ross

[-- Attachment #2: Type: text/html, Size: 1415 bytes --]