linux-denx missing the needed hash

linux-denx missing the needed hash

[Otavio Salvador]

Hello Marek,

(added meta-freescale and WD in Cc)

Using our builder we found that linux-denx seems to have been rebased
and no tag made for the previous hash. So users cannot fetch the hash
that is pointed in the recipe. You can see it in the log below:

WARNING: Failed to fetch URL
git://git.denx.de/linux-denx.git;branch=denx/DENX-v3.9-stable,
attempting MIRRORS if available
ERROR: Fetcher failure: Unable to find revision
7c75b82904fa555ce7988b97619b85a436a8ed12 in branch
denx/DENX-v3.9-stable even from upstream
ERROR: Function failed: Fetcher failure for URL:
'git://git.denx.de/linux-denx.git;branch=denx/DENX-v3.9-stable'.
Unable to fetch URL from any source.
ERROR: Logfile of failure stored in:
.../build/tmp/work/m53evk-poky-linux-gnueabi/linux-denx/3.9-r0/temp/log.do_fetch.25838

Could you address this with DENX?

Regards,

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: linux-denx missing the needed hash

[Wolfgang Denk]

Dear Otavio,

In message <CAP9ODKpUd30kbA2m4Qrd-Yk5sG=HoMz3x-DkyugSVXK=rQ7neQ@mail.gmail.com> you wrote:
> 
> Using our builder we found that linux-denx seems to have been rebased
> and no tag made for the previous hash. So users cannot fetch the hash
> that is pointed in the recipe. You can see it in the log below:

Your conclusions are wrong.  Nothing has changed in our repository.
The build error results from your own commit [1] where you added
incorrect (and obviously untested) code.


[1] https://github.com/Freescale/meta-fsl-arm-extra/commit/04ea0ff09376af9275c3587e6cd947c2a94738fa


Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de
A wise person makes his  own  decisions,  a  weak  one  obeys  public
opinion.                                           -- Chinese proverb

Re: linux-denx missing the needed hash

[Otavio Salvador]

On Wed, Feb 19, 2014 at 2:17 PM, Wolfgang Denk <wd@denx.de> wrote:
> Dear Otavio,
>
> In message <CAP9ODKpUd30kbA2m4Qrd-Yk5sG=HoMz3x-DkyugSVXK=rQ7neQ@mail.gmail.com> you wrote:
>>
>> Using our builder we found that linux-denx seems to have been rebased
>> and no tag made for the previous hash. So users cannot fetch the hash
>> that is pointed in the recipe. You can see it in the log below:
>
> Your conclusions are wrong.  Nothing has changed in our repository.
> The build error results from your own commit [1] where you added
> incorrect (and obviously untested) code.
>
>
> [1] https://github.com/Freescale/meta-fsl-arm-extra/commit/04ea0ff09376af9275c3587e6cd947c2a94738fa

I think I tested this when added the branch information. It seems it
has a tag pointing to the hash and I will check how to use it (BitBake
has some issues to deal with the tags) but it would be nice if you or
someone else at DENX might help on this as I don't own a m53evk board.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: linux-denx missing the needed hash

[Wolfgang Denk]

Dear Otavio,

In message <CAP9ODKr50h9hFFPRkeLO8iKkextuyNoxaBMS-nHc=5ZvKbnBJw@mail.gmail.com> you wrote:
>
> > Your conclusions are wrong.  Nothing has changed in our repository.
> > The build error results from your own commit [1] where you added
> > incorrect (and obviously untested) code.
...
> I think I tested this when added the branch information. It seems it

I don't think so.  The commit does not move around though different
branches.

> has a tag pointing to the hash and I will check how to use it (BitBake
> has some issues to deal with the tags) but it would be nice if you or
> someone else at DENX might help on this as I don't own a m53evk board.

I don't see why you would actually use any branch or tag information.
As you just learned, this is not reliable.  Just refer to the commit
ID, and you are all fine.

As for help: I cannot see which help you would need to run a test
build.  You can check yourself if it builds or fails.  Remember, the
code was orking before you broke it.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de
"Engineering without management is art."               - Jeff Johnson

Re: linux-denx missing the needed hash

[Otavio Salvador]

Hello Wolfgang,

On Wed, Feb 19, 2014 at 2:30 PM, Wolfgang Denk <wd@denx.de> wrote:
> Dear Otavio,
>
> In message <CAP9ODKr50h9hFFPRkeLO8iKkextuyNoxaBMS-nHc=5ZvKbnBJw@mail.gmail.com> you wrote:
>>
>> > Your conclusions are wrong.  Nothing has changed in our repository.
>> > The build error results from your own commit [1] where you added
>> > incorrect (and obviously untested) code.
> ...
>> I think I tested this when added the branch information. It seems it
>
> I don't think so.  The commit does not move around though different
> branches.
>
>> has a tag pointing to the hash and I will check how to use it (BitBake
>> has some issues to deal with the tags) but it would be nice if you or
>> someone else at DENX might help on this as I don't own a m53evk board.
>
> I don't see why you would actually use any branch or tag information.
> As you just learned, this is not reliable.  Just refer to the commit
> ID, and you are all fine.

This is required in current BitBake fetcher code and thus in master.
This is new in 1.6 release.

> As for help: I cannot see which help you would need to run a test
> build.  You can check yourself if it builds or fails.  Remember, the
> code was orking before you broke it.

Sorry Wolfgang but you are wrong. As I state before this is need in master.

Help? I do think someone supporting and doing regular tests would be
good. If you were subscribed to the mailing list you'd know I sent
another fix to make m53evk to build in Dora when using X11 which were
failing and no one found it.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: linux-denx missing the needed hash

[Wolfgang Denk]

Dear Otavio,

In message <CAP9ODKq5gBT3sm7XmgyiwkYqcTxUPt_VsJfaCMZPMCu3RPoCrQ@mail.gmail.com> you wrote:
> 
> > I don't see why you would actually use any branch or tag information.
> > As you just learned, this is not reliable.  Just refer to the commit
> > ID, and you are all fine.
> 
> This is required in current BitBake fetcher code and thus in master.

The build fails because you selected the wrong branch.

Commit 7c75b82 (tag: m53evk-2013-05-02-v3.9, branch:
m53evk-rel-2013-05-02-v3.9) was checked in on Thu Apr 4 17:38:54 2013.
Neither the commit nor the branch has have ever changed since they
were added.

If you want to use branch names, just use the correct ones.  If in
doubt, just ask git:

	-> git branch --contains 7c75b82904fa555ce7988b97619b85a436a8ed12
	  m53evk-rel-2013-05-02-v3.9

I cannot see why you have such a problem just to fix that name?

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de
I perceive a possibility of an immediate  chronological  sequence  of
events which includes a violence.
                        - Terry Pratchett, _The Dark Side of the Sun_

Re: linux-denx missing the needed hash

[Otavio Salvador]

On Wed, Feb 19, 2014 at 4:21 PM, Wolfgang Denk <wd@denx.de> wrote:
> In message <CAP9ODKq5gBT3sm7XmgyiwkYqcTxUPt_VsJfaCMZPMCu3RPoCrQ@mail.gmail.com> you wrote:
>>
>> > I don't see why you would actually use any branch or tag information.
>> > As you just learned, this is not reliable.  Just refer to the commit
>> > ID, and you are all fine.
>>
>> This is required in current BitBake fetcher code and thus in master.
>
> The build fails because you selected the wrong branch.
>
> Commit 7c75b82 (tag: m53evk-2013-05-02-v3.9, branch:
> m53evk-rel-2013-05-02-v3.9) was checked in on Thu Apr 4 17:38:54 2013.
> Neither the commit nor the branch has have ever changed since they
> were added.
>
> If you want to use branch names, just use the correct ones.  If in
> doubt, just ask git:
>
>         -> git branch --contains 7c75b82904fa555ce7988b97619b85a436a8ed12
>           m53evk-rel-2013-05-02-v3.9

I know git command line.

> I cannot see why you have such a problem just to fix that name?

I am fixing; my main problem is no one supporting those board and
doing tests with it.

As you can see, after Marek sending the board for inclusion, most
fixes (even for build failures) has been done by me:

commit b43b847062c211a497e48d6443137a7caa184463
Author: Otavio Salvador <otavio@ossystems.com.br>
Date:   Wed Feb 19 10:15:04 2014 -0300

    m53evk: Fix build failure due non-FSL kernel use with AMD GPU

    The m53evk board uses the DENX kernel which is based in mainline and
    does not has the need drivers for AMD GPU driver to build. This
    patches makes it use the Framebuffer driver instead fixing a build
    failure found in our autobuilder.

    Change-Id: Iad1dfbb7f9ed342f374d9721228eff09e821377f
    Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
    Cc: Marek Vasut <marex@denx.de>

commit 3d1464bed2c6fefe235f9427c33642e7dc964153
Author: Otavio Salvador <otavio@ossystems.com.br>
Date:   Fri Nov 1 19:17:47 2013 -0200

    m53evk: Do not enable GStreamer plugin

    The kernel used for this board is based in the mainline kernel and
    does not provide the Freescale specific headers so we cannot build the
    GStreamer specific plugin.

    Change-Id: I382759bc05a74c638cb80dd61ce950bf896d33cf
    Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>

commit 78410703f94ff03a03959f204b3febcfb1ca1ebe
Author: Otavio Salvador <otavio@ossystems.com.br>
Date:   Fri Nov 1 19:16:15 2013 -0200

    m28evk: Do not enable GStreamer plugin

    The kernel used for this board is based in the mainline kernel and
    does not provide the Freescale specific headers so we cannot build the
    GStreamer specific plugin.

    Change-Id: Icd9e08c3e774ea3a0df2c532ef52619d378a4fca
    Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>

commit 0512f5fd03033b8ab707fcc8624fac68b79a321d
Author: Gary Thomas <gary@mlbassoc.com>
Date:   Thu Oct 31 14:59:29 2013 +0000

    machine/*: Allow kernel provider override

    All machine configurations now allow user override of the preferred
    kernel provider.

    Signed-off-by: Gary Thomas <gary@mlbassoc.com>

commit 53cae59d3a7649052c7ac9d698f82367a7970e70
Author: Otavio Salvador <otavio@ossystems.com.br>
Date:   Sat Aug 17 18:55:56 2013 -0300

    Rework machines to use dtb file in KERNEL_DEVICETREE variable

    The linux-dtb.inc has been reworked to use the Linux kernel build
    system to generate the DeviceTree binaries so now KERNEL_DEVICETREE
    variable should has the /target/ name, not the file path. This patch
    reworks following machines:

     - imx233-olinuxino-*
     - m28evk
     - m53evk

    Change-Id: I824e818f50de094fe1b25ef3ea202abd90093650
    Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>

commit f6cc86276bb7192c1f23dda8b90f2f5cf1a52ff4
Author: Otavio Salvador <otavio@ossystems.com.br>
Date:   Wed Jul 24 13:53:14 2013 -0300

    m53evk: Remove duplicated setting of U-Boot preferred provider

    The system already default for u-boot-fslc as provider so we shouldn't
    hardcode it here.

    Change-Id: I4472561b74f857813b5f1f2b9d8ed2d68f479c8d
    Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>

commit 7995948391a9052e4c83a9599af7dd8b6beb091a
Author: Marek Vasut <marex@denx.de>
Date:   Tue Jul 2 01:49:07 2013 +0000

    m53evk: Add DENX M53EVK support

    Add DENX M53EVK board configuration file. This uses the linux-denx
    kernel repository.

    Change-Id: I4c7ad1446a3bf1cd5c149701861ed745025d1449
    Signed-off-by: Marek Vasut <marex@denx.de>

commit 08348ddf96f2beafa1522877c53148e714c03cba
Author: Otavio Salvador <otavio@ossystems.com.br>
Date:   Thu May 16 11:35:29 2013 -0300

    m28evk: Add DENX M28EVK board support

    Change-Id: I60012955c249d0a22726ef5127dc1cfbc5cfdd50
    Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>

Wolfgang, please try to get someone watching those boards (and they
are still marked as orphan by the way).

If you don't see value in those being supported better to remove them.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: linux-denx missing the needed hash

[Wolfgang Denk]

Dear Otavio,

In message <CAP9ODKoERSkfd3QaBgK-r_n6PaB+gXQA7apARRdCt0ywnPuUPA@mail.gmail.com> you wrote:
>
> I am fixing; my main problem is no one supporting those board and
> doing tests with it.
> 
> As you can see, after Marek sending the board for inclusion, most
> fixes (even for build failures) has been done by me:

But... aren't those fixes only needed due to changes you made
elsewhere?

>     m53evk: Fix build failure due non-FSL kernel use with AMD GPU
> 
>     The m53evk board uses the DENX kernel which is based in mainline and
>     does not has the need drivers for AMD GPU driver to build. This
>     patches makes it use the Framebuffer driver instead fixing a build
>     failure found in our autobuilder.

It worked when Marek submitted it, right?  So something else was
changed?

>     m53evk: Do not enable GStreamer plugin
> 
>     The kernel used for this board is based in the mainline kernel and
>     does not provide the Freescale specific headers so we cannot build the
>     GStreamer specific plugin.

This sounds as if it was a Bad Thing to use mainline code.  This is an
attitude I do not understand.  Is it not that out-of-tree code that
breaks things?  So if changes are needed, should the focus not be on
getting rid of the out-of-tree code - either by pushing it upstream,
or, if it is not acceptable for mainline, by dropping it? It may make
sense to use out-of-tree code in certain configurations, but to make
it mandatory for all looks the wrong approach to me.

> Wolfgang, please try to get someone watching those boards (and they
> are still marked as orphan by the way).
> 
> If you don't see value in those being supported better to remove them.

But this is not the problem. The problem is that you added a modifi-
cation which introduced a bug, and then complain, and blame others
for the resulting build breakage.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de
"Plan to throw one away. You will anyway."
                              - Fred Brooks, "The Mythical Man Month"

Re: linux-denx missing the needed hash

[Otavio Salvador]

Hello Wolfgang,

On Wed, Feb 19, 2014 at 7:43 PM, Wolfgang Denk <wd@denx.de> wrote:
> Dear Otavio,
>
> In message <CAP9ODKoERSkfd3QaBgK-r_n6PaB+gXQA7apARRdCt0ywnPuUPA@mail.gmail.com> you wrote:
>>
>> I am fixing; my main problem is no one supporting those board and
>> doing tests with it.
>>
>> As you can see, after Marek sending the board for inclusion, most
>> fixes (even for build failures) has been done by me:
>
> But... aren't those fixes only needed due to changes you made
> elsewhere?

Not all; I won't enumerate them and you can check the logs.

>>     m53evk: Fix build failure due non-FSL kernel use with AMD GPU
>>
>>     The m53evk board uses the DENX kernel which is based in mainline and
>>     does not has the need drivers for AMD GPU driver to build. This
>>     patches makes it use the Framebuffer driver instead fixing a build
>>     failure found in our autobuilder.
>
> It worked when Marek submitted it, right?  So something else was
> changed?

If you read the looks it didn't. The problem is that Marek didn't
build X11 and he didn't figure it out.

>>     m53evk: Do not enable GStreamer plugin
>>
>>     The kernel used for this board is based in the mainline kernel and
>>     does not provide the Freescale specific headers so we cannot build the
>>     GStreamer specific plugin.
>
> This sounds as if it was a Bad Thing to use mainline code.  This is an
> attitude I do not understand.  Is it not that out-of-tree code that
> breaks things?  So if changes are needed, should the focus not be on
> getting rid of the out-of-tree code - either by pushing it upstream,
> or, if it is not acceptable for mainline, by dropping it? It may make
> sense to use out-of-tree code in certain configurations, but to make
> it mandatory for all looks the wrong approach to me.

Subscribe the list and review the patches. It is a little too late to
complain about a commit log already merged.

>> Wolfgang, please try to get someone watching those boards (and they
>> are still marked as orphan by the way).
>>
>> If you don't see value in those being supported better to remove them.
>
> But this is not the problem. The problem is that you added a modifi-
> cation which introduced a bug, and then complain, and blame others
> for the resulting build breakage.

Wolfgang, the problem is you are trying to find someone to point as guilt.

Even if I did a mistake in the branch name (which I am sure it worked
at that time) I don't make money selling M28EVK or M53EVK, you do. So
who should be keeping an eye if boards are working in the BSP?

As I said before:

>> I am fixing; my main problem is no one supporting those board and
>> doing tests with it.

I pushed to fix already so DENX boards are building. If you don't see
value in having the boards in the BSP those can be dropped from the
BSP. I would prefer if you, Marek or someone else could keep an eye on
those and do regular tests so we catch problems.

I personally told Marek about the X11 issue more than a week ago. It
ended I fixing it ... please stop and think if it is right.

By the way, this is my last e-mail in this thread.

Have a good day.

Regards,

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: linux-denx missing the needed hash

[Wolfgang Denk]

Dear Otavio,

In message <CAP9ODKpExT9=XNe0ZDh-S0amFh0rGqmctuuOq0G3MzJHm2vYqw@mail.gmail.com> you wrote:
> 
> > It worked when Marek submitted it, right?  So something else was
> > changed?
> 
> If you read the looks it didn't. The problem is that Marek didn't
> build X11 and he didn't figure it out.

Yes, but then: we are not interestedn in supporting X11 on this
platform.  All users we are aware of prefer to use QtE when they need
a GUI.  So simply don't build X11.

> Subscribe the list and review the patches. It is a little too late to
> complain about a commit log already merged.

Thanks, but actually I am not interested in out-of-tree code.

> Even if I did a mistake in the branch name (which I am sure it worked
> at that time) I don't make money selling M28EVK or M53EVK, you do. So
> who should be keeping an eye if boards are working in the BSP?

I am not aware of any of our customers who ever tried the fsl-extre
stuff.  We have to provide support to our customers, and supporting
out-of-tree code is insane from a commercial point of view.  This is
why we only support them with mainline code.

But you sidetrack again: in community work, it is an established rule
that he who breakes working code is normally also responsible to fix
it.  Here it is a trivial build error, where no knowledge about the
actual hardware nor any actual testing on the board is needed, so I
really cannot see why you make such a fuss.

> I pushed to fix already so DENX boards are building. If you don't see
> value in having the boards in the BSP those can be dropped from the
> BSP. I would prefer if you, Marek or someone else could keep an eye on
> those and do regular tests so we catch problems.

You are suposed to test your commits so that they do not break the
build.

> I personally told Marek about the X11 issue more than a week ago. It
> ended I fixing it ... please stop and think if it is right.

I think it is.

- We are not much interested in X11 for this platform.  None of our
  customers and known users uses it.  Why should we invest efforts in
  testing or fixing unused features?

- It appears the breakage comes from out-of-tree code.  All the Yocto
  defined images build and run fine in a plain Yocto environment.
  This applies fully to Yocto 1.5.1 (see for example our ELDK v5.5
  which is based on this), and it applies mostly to Yocto 1.6_M2
  (we still have build issues there, but these are related to the
  Xenomai integration an not to using a mainline kernel [without
  referencing a branch name, btw.] or X11 or gstreamer or such].
  
  Especially with a commercial background in we need code that can be
  maintained over a long period of time - at least 5...10 years, maybe
  more.  This simply doesn't work with out-of-tree code, so we have no
  interest in fixing out-of-tree code, or problems caused by it.

> By the way, this is my last e-mail in this thread.

Agreed.  There have not been really new arguments, so we can stop here.

Thanks.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de
If you're out of tree, you don't exist.
     - David Woodhouse in <1304620350.2398.29.camel@i7.infradead.org>

Re: linux-denx missing the needed hash

[Kumar, Dheeraj]

[-- Attachment #1: Type: text/plain, Size: 704 bytes --]

Hello Guys,

I am trying to build the package and it is not even downloading from the repo. It get the same error:


WARNING: Failed to fetch URL

git://git.denx.de/linux-denx.git;branch=denx/DENX-v3.9-stable,

attempting MIRRORS if available

ERROR: Fetcher failure: Unable to find revision

7c75b82904fa555ce7988b97619b85a436a8ed12 in branch

denx/DENX-v3.9-stable even from upstream

ERROR: Function failed: Fetcher failure for URL:

'git://git.denx.de/linux-denx.git;branch=denx/DENX-v3.9-stable'.

Unable to fetch URL from any source.

seems like current version of bitbake requires hash and tags as well to use with it.
Any suggestions how to proceed ?

Thanks,
Dheeraj

[-- Attachment #2: Type: text/html, Size: 3663 bytes --]