[oe] SRCPV issue? Was: [PATCH 3/3] linux-openmoko: rename recipes to common PN_PV scheme and adjust om-gta* machine configs

[Martin Jansa]

On Wed, May 26, 2010 at 2:20 PM, Martin Jansa <martin.jansa at gmail.com> wrote:
> On Tue, May 25, 2010 at 4:37 PM, Khem Raj <raj.khem at gmail.com> wrote:
>> On (25/05/10 15:30), Martin Jansa wrote:
>>> * unify preferred provider and version for om-gta02
>>>
>>> Signed-off-by: Martin Jansa <Martin.Jansa at gmail.com>
>>
>> Acked-by: Khem Raj <raj.khem at gmail.com>
>>
>>> ---
>>>  conf/distro/openmoko.conf                          |    8 -
>>>  conf/machine/om-3d7k.conf                          |    3 +-
>>>  conf/machine/om-gta01.conf                         |   12 +--
>>>  conf/machine/om-gta02.conf                         |   10 +-
>>>  recipes/linux/linux-openmoko-2.6.24_git.bb         |   28 ----
>>>  recipes/linux/linux-openmoko-2.6.28_git.bb         |   28 ----
>>>  .../0007-Enable-UBI-UBIFS.patch                    |  149 ++++++++++++++++++++
>>>  .../linux/linux-openmoko-2.6.29/fix-install.patch  |   23 +++
>>>  recipes/linux/linux-openmoko-2.6.32_git.bb         |   46 ------
>>>  recipes/linux/linux-openmoko-2.6.34_git.bb         |   39 -----
>>>  .../0007-Enable-UBI-UBIFS.patch                    |  149 --------------------
>>>  .../linux-openmoko-shr-devel/fix-install.patch     |   23 ---
>>>  recipes/linux/linux-openmoko-shr-devel_git.bb      |   26 ----
>>>  recipes/linux/linux-openmoko_2.6.24.bb             |   28 ++++
>>>  recipes/linux/linux-openmoko_2.6.28.bb             |   28 ++++
>>>  recipes/linux/linux-openmoko_2.6.29.bb             |   26 ++++
>>>  recipes/linux/linux-openmoko_2.6.32.bb             |   46 ++++++
>>>  recipes/linux/linux-openmoko_2.6.34_git            |   39 +++++
>>>  18 files changed, 347 insertions(+), 364 deletions(-)
>>>  delete mode 100644 recipes/linux/linux-openmoko-2.6.24_git.bb
>>>  delete mode 100644 recipes/linux/linux-openmoko-2.6.28_git.bb
>>>  create mode 100644 recipes/linux/linux-openmoko-2.6.29/0007-Enable-UBI-UBIFS.patch
>>>  create mode 100644 recipes/linux/linux-openmoko-2.6.29/fix-install.patch
>>>  delete mode 100644 recipes/linux/linux-openmoko-2.6.32_git.bb
>>>  delete mode 100644 recipes/linux/linux-openmoko-2.6.34_git.bb
>>>  delete mode 100644 recipes/linux/linux-openmoko-shr-devel/0007-Enable-UBI-UBIFS.patch
>>>  delete mode 100644 recipes/linux/linux-openmoko-shr-devel/fix-install.patch
>>>  delete mode 100644 recipes/linux/linux-openmoko-shr-devel_git.bb
>>>  create mode 100644 recipes/linux/linux-openmoko_2.6.24.bb
>>>  create mode 100644 recipes/linux/linux-openmoko_2.6.28.bb
>>>  create mode 100644 recipes/linux/linux-openmoko_2.6.29.bb
>>>  create mode 100644 recipes/linux/linux-openmoko_2.6.32.bb
>>>  create mode 100644 recipes/linux/linux-openmoko_2.6.34_git

Changed subject to call other devs attention to this issue (who are
not interested in linux-openmoko).

> Too late I found flaw of this (or in how SRCPV works) :/.
>
> As long as those were recipes from same GIT repo but with different PN
> it was ok, because in persistent cache they got different keys:
>
> sqlite> select * from BB_URI_LOCALCOUNT where key like '%linux-openmoko%';
> git:git.openmoko.org.git.kernel.git-linux-openmoko-shr-devel_rev|973a41fce60e5f6edfcf0eaf94056a57eee65e16
> git:git.openmoko.org.git.kernel.git-linux-openmoko-shr-devel_count|1
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.32_rev|a9254be10ac2294ea20165a87c09ea6afcf66d94
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.32_count|0
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.31_rev|dd32737aa524e09c1d141a0e735dd58674552244
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.31_count|0
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.28_rev|8c65792a5c83c76d662a617a7c4e1ae8104bb6a5
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.28_count|0
> git:git.openmoko.org.git.kernel.git-linux-openmoko-shr-drm-devel_rev|e62a4ae1c6783f41b41a9ac3d258786586b65a40
> git:git.openmoko.org.git.kernel.git-linux-openmoko-shr-drm-devel_count|0
> git:git.openmoko.org.git.kernel.git-linux-openmoko-devel_rev|a15608f241a40b41fed5bffe511355c2067c4e88
> git:git.openmoko.org.git.kernel.git-linux-openmoko-devel_count|0
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.24_rev|fb42ce6724576fc173faf8abfb04aa2c36d213b7
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.24_count|0
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.34_rev|dd1225cc08c3375bf80289ac1965c724881b149a
> git:git.openmoko.org.git.kernel.git-linux-openmoko-2.6.34_count|1
>
> now they all use
> 'git:git.openmoko.org.git.kernel.git-linux-openmoko_rev' and each
> recipe parsing seems to increase _count by number of
> linux-openmoko*.bb recipes (because _rev is changed between each
> recipe) :/.
>
> git:git.openmoko.org.git.kernel.git-linux-openmoko_rev|dd1225cc08c3375bf80289ac1965c724881b149a
> git:git.openmoko.org.git.kernel.git-linux-openmoko_count|78
>
> Please advise what should be best fix for this. I'm OK with reverting
> my last 2 commits and keeping unique PN for them, but maybe the same
> issue can be found in other recipes too.
>
> Using recipe name for persistent cache could be easy solution but
> needs updated bitbake :/.

Ah not easy at all, we want persitent cache key as stable as possible
(renaming recipe while keeping same PN/PV should stay on upgradable
path with next SRCREV change - needs same key). Appending PV here
(before expanding as expanded PV usually contains SRCREV) will be
maybe better. But both "solutions" need also to migrate existing cache
keys, which is difficult and error-prone.

Partial revert seems like best solution now :/.

Regards