Age | Commit message (Collapse) | Author | Files | Lines |
|
Change-Id: I86f524f656901352a0de9484b10776ddd17ef2d0
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
To avoid further issues with sub-libraries bumping ABIs one by one
we adapted to the MAJOR_ABI config feature.
That shall allow us to:
- co-install old and new binaries
- avoid recursive ldd resolving to the wrong (old vs new ABI) libs
- all libs being directly recognizable which code base they are from
Change-Id: I14c7bd1749969a812be9730829ba009401d443a2
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This is switching the existing binary packages to the new versioning
being major ABI (currently 17.05) for all of the packages.
Change-Id: If01167adc77aacf1de9d538ff55b508c28342185
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Now that we are swicthign to MAJOR_ABI builds we will have no
conflicting sublibrary versions anymore, drop the old crutch.
Change-Id: I54fa90ee6b5709ff6b2000fefbfba4ea229b2bf9
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I734282387433b2218b2481c69c39ae0ad2b8537e
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
* Add update-helper-control.py script to easen rename packages for new
DPDK versions.
- This script from libboost helps convert all package names
from the old ABI name to the new one.
- thanks to Jan Blunck for the conversion from boost!
- the script was later updated to be python3 ready and fixing the script
for all warnings/errors (but not all info) thrown by python code checkers.
* Add update-helper-symbols.sh script to easen modifying the package and
soname references in symbols files for new DPDK versions.
- We still want to check "real" symbol delta on any update, but to make
dpkg-gensymbols find the new libs the symbols files need to be updated.
Change-Id: I73a454b8fe628c05569e689f5c14f3b73d2ec4cc
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Initial 17.05 merge git changelog entry which will be extended by new
libs features, fixes as needed along the merge. So far only
mentioning the switch to use MAJOR_ABI.
Change-Id: I7d4c8d85510ab5889221b813ac20e415a8638ce6
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This script from libboost helps convert all package names from the
old ABI name to the new one - thanks to Jan Blunck for the initial
conversion!
Change-Id: I6c74fab9f3122cae464ac0bc705bfb32c9f1227d
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
with Debian dir b312c5b3fb7ca9fa049370b932f7d64c929ef0eb
|
|
|
|
Change-Id: I99549a49fc9a405ce42017c2f4b08b68c379169e
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I8b6e5ba6cdcba5bf040b5c9b0f77ffb0db3bd9af
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
debian/copyright flagged the whole lib/librte_eal/linuxapp/ content
as distributed under GPL2, but it's only the kernel modules. Update
the copyright file to clarify this.
Change-Id: I891a798f3a48b3556a8b6d547f790b2d4c4c55f6
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
Change-Id: I657970c3c60bf2e062a3862a95e7e39da24a7d66
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I83d6b19ceab5fdac544b149aff78183ef8a45ff6
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I1339b4368deb0e745a2df11c0314c659cb7b9a6e
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
This is a workaround for the assumption of install-kmod functionality
that the running kernel is identical with the kernel we build against.
This fixes commit 75964339a3 ("Make debian/rules file call install-kmod").
Change-Id: If5b9a99db36d4a9130ee9876ef93d581132369fc
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
Change-Id: Ie769436ef8de824ad85ac5fd10f12de4e6f8f970
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
Change-Id: Ie619ee1992fd6d9d41e0b427707a825658c4dfc9
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
Lets pass the multiarch libdir at install time so that the links embedded
in the dpdk-dev package (build environment) are correct.
Change-Id: I4f907ad88721f2155a300dc6ad6724a686b7b194
Signed-off-by: Jan Blunck <jblunck@infradead.org>
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
Change-Id: I37a1382a4c8b72665032f9410609b9eabc34291f
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: Ib15a339c8050bb0594ce8333133e923771d2fc57
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I1e965265578efaaf08e5628607f53d2386d2df9f
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I4a19ade7403c6b8b923f10e76c227c392594e698
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
The cross dependencies in DPDK libraries would force us to conflict
with older ABI versions to avoid double ld.so resolution
(https://lists.ubuntu.com/archives/ubuntu-devel/2017-February/039659.html)
The conflicts would make the old versions and thereby the dependent
applications non-installable until there is a rebuild (as other updates
would do).
But for our case that would deny most benefits of version named
libraries.
Instead of adding a conflicts - whenever possible - provide a backward
compatible wrapper symlinking to the new (ABI compatible) implementation.
Change-Id: If9fa9cb9474ad0ac555bdf1a0c72f701ddaf2926
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This was wrong, remove it from our master.
When backporting to Ubuntu/Debian this change to the changelog will not
be a drop from the log, but instead a change made in e.g. 16.11-2 - so
be careful when picking.
Change-Id: I226d6454b8e04bfd35cf8b275b6f1894fc503e76
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
The old approach was just wrong, we have two cases:
1. a distribution without the old sonames, no transition needed
2. a distribution with the old sonames
- continue to provide them by the old binary packages works fine
(they are named libs for a reason)
- provide new ones in new binary packages built now
- whenever anything is rebuilt it picks up the new dependency
The worst is that I did this :-/ not thinking at all it seems.
Not hurtin on case #1, but a lot on case #2 - you likely want an
archive admin to drop the transitional packages after binaries with this
fix are built.
Change-Id: Ie60245818bc55a075a8b565dd38614351811e849
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: If4643c961ef34d79aa3c1e3a7ff6e7133d7e5f42
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I8677177124b89641854187a88cb876c4594ea8d7
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I61956681fd36f3ecd223271b1538b57e0d955a01
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This test part of the test was arch specific e.g. not xen is
not build on ppc64el.
Change-Id: I71bed3a353c34b2785791406200eaf9bb275db2a
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This dependency was arch specific so the test failed on e.g. ppc64el
Change-Id: I162bc77288aae3be56e1840764914d1264786b98
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This dependency is only needed for the rather uncommon dpdk-pmdinfo
which is more a developer tool these days still.
To soften the relationship helps the joint maintenance with Ubuntu.
Change-Id: I50dd2ab6af3a615249112416a442c589a2cdf4a6
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
To ease maintenance appyling wrap-and-sort to Debian control.
See man wrap-and-sort for details.
Change-Id: I961d336ac533b71295403462c8c83071f2f6eb2e
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: Ib1ed056632dc4d46de43eb7f8d9fc7343d8f6b60
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: Idd522f11cf20b6321ccd1c7db949f5fcb10f82d3
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: Ia068e67d8885ef82efdd3df339f7f05aa6955008
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
Change-Id: I21f9cb9c0ba254de5ba7406cdec0c3bdfb05e60e
Signed-off-by: Luca Boccassi <luca.boccassi@gmail.com>
|
|
Change-Id: Ic775201177d138e4b58a59ffa5c030676347efcf
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
For now the autotest is too unreliable, so we run it to get some
logs in different environments, but never (want to ) fail until
it is stable.
This was already the intention and worked if certain tests failed
as we didn't check on the RC of the tests, but if the test overall
fatally fails we had a FAIL due to the set -e.
Change-Id: I54052351113a974d3e7810d55405f063e1fbe16a
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I665c000c306fe438d57b3b2d04bf4d531d9dcc61
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I87c435b2c77e704c7d3fb84fe722931f8b18d0aa
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I097e42aacdd75f7e38137754b17d938b7053b07d
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I6335fc54749b0510642733dca8c59fd8f14aebc7
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I98ca53db88eac6c8ebd948d4700fe7210165674c
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This is a bit of a "decision" we don't have many (in Debian actually
none) depending packages so far.
In a library that would behave we would ship dpdk-16.11 as new source
along dpdk-16.07, but as discussed we don't want to do that just yet.
OTOH the version 5 is fully backward compatible with version 4, so for
now we can handle it like a package rename transition.
On top of the api bump the library got renamed by upstream from
libethdev to librte-ethdev to follow their usual naming scheme.
That implies to rebuild dependent packages in distributions that have
them already (in Ubuntu) as a transitional joint package upload.
Change-Id: I13e0fc99ac1adb230f1cfc7491eb7cf162996df3
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I96408e071b5b57faedd58e56274017b1088a8ba7
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This is a bit of a "decision" we don't have many (in Debian actually
none) depending packages so far.
In a library that would behave we would ship dpdk-16.11 as new source
along dpdk-16.07, but as discussed we don't want to do that just yet.
OTOH the version 2 is fully backward compatible with version 1, so for
now we can handle it like a package rename transition.
That implies to rebuild dependent packages in distributions that have
them already (in Ubuntu) as a transitional joint package upload.
Change-Id: I028c24f6cc657731b1d3f058bd484fdb5ebacb49
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
Change-Id: I70cbbc8f23650b77abb1f8134b3190d4de2058d1
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|
|
This is a bit of a "decision" we don't have many (in Debian actually
none) depending packages so far.
In a library that would behave we would ship dpdk-16.11 as new source
along dpdk-16.07, but as discussed we don't want to do that just yet.
OTOH the version 3 is fully backward compatible with version 2, so for
now we can handle it like a package rename transition.
That implies to rebuild dependent packages in distributions that have
them already (Ubuntu) as a transitional package upload.
Change-Id: I92e79f206a0bd752d775584c88ea2559c652b606
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
|