| Age | Commit message (Collapse) | Author |
|---|
|
The sstate-cache for libecj-bootstrap is not shareable across multiple
machines because the parallelisation configuration of Bitbake is being
embedded into the sstate hash of libecj-bootstrap.
Exclude BB_NUMBER_THREADS from the do_compile sstate hashing to avoid
this issue.
Commit message adapted from a5257d41f32812bc9426183cc8d2b6f07617b0d5.
Signed-off-by: Wes Lindauer <wesley.lindauer@gmail.com>
Signed-off-by: Richard Leitner <richard.leitner@skidata.com>
|
|
Fixes QA Issue: classpath-native: native/nativesdk class is not inherited last
Signed-off-by: Douglas Royds <douglas.royds@taitradio.com>
Signed-off-by: Richard Leitner <richard.leitner@skidata.com>
|
|
Short descriptions should go into SUMMARY (DESCRIPTION will
get the same value if not set.)
Signed-off-by: André Draszik <andre.draszik@jci.com>
Tested-by: Richard Leitner <richard.leitner@skidata.com>
Signed-off-by: Richard Leitner <richard.leitner@skidata.com>
|
|
Piping 'find' output into multiple files to re-read
them seems inelegant and is error prone - just use a
pipe with appropriate options instead.
This avoids potential problems with funny file names,
and now also makes use of BB_NUMBER_THREADS to speed
up compilation.
This is a better example to copy from now...
Signed-off-by: André Draszik <andre.draszik@jci.com>
Tested-by: Richard Leitner <richard.leitner@skidata.com>
Signed-off-by: Richard Leitner <richard.leitner@skidata.com>
|
|
The OE-Core now warns when someone uses ${COREBASE}/LICENSE so we
should move to ${COMMON_LICENSE_DIR}/MIT to fix this deprecated use.
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
|
|
Signed-off-by: Adrian Calianu <adrian.calianu@enea.com>
Signed-off-by: Maxin B. John <maxin.john@intel.com>
|
|
bitbake rev 67a7b8b02 "build: don't use $B as the default cwd for
functions" (included in current bitbake master) breaks the assumption
that do_unpackpost runs inside the build directory. Now that has to be
set explicitly, which is also okay for older bitbake versions.
Signed-off-by: Patrick Ohly <patrick.ohly@intel.com>
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
|
|
Various paths to files are hardcoded in either helper scripts or the
resulting binaries (like path to Java class libraries). While this works
well when using the files on the build machine, things start breaking
down when the sstate gets shared between multiple workers in CI, as
the paths between workers can differ.
Instead of using these hardcoded paths, form the paths at runtime and pass them as parameters when required.
Signed-off-by: Erkka Kääriä <erkka.kaaria@intel.com>
Signed-off-by: Maxin B. John <maxin.john@intel.com>
|
|
We had:
virtual/jar-native provided by fastjar-native
virtual/java-native provided by cacao-native and jamvm-native
virtual/javac-native provided by ecj-bootstrap-native
Then openjdk-8-native added and also provides:
virtual/java-native and virtual/javac-native
(it also installs ${STAGING_BINDIR_NATIVE}/jar but
doesn't add virtual/jar-native in the PROVIDES)
which introduces circular dependencies,
java-bootstrap-components.bbclass is to avoid the
circular dependencies but it's not a correct way
since it make two conflicted packages build into
the same sysroot:
DEPENDS_prepend_class-native = " ecj-bootstrap-native "
DEPENDS_prepend_class-target = " virtual/javac-native "
virtual/javac-native is provided by openjdk-8-native,
and it assumed that ecj-bootstrap-native, fastjar-native and
cacao-native/jamvm-native are always built before openjdk-8-native,
then in the do_install of penjdk-8-native, it removes them from
staging dirs first, then install (create symlinks for) these files
in staging dirs directly, which results with empty sstate and
builds from sstate cache or builds not in this assumption
will fail.
The fixes here includes:
- Revert "meta-java: rely on well known bootstrap-path" in
previous commit to remove the improper
java-bootstrap-components.bbclass
- Do not let openjdk-8-native provides jar, java and javac,
just like what we do in icedtea7-native, which can provides
but not to avoid circular dependencies and conflicts.
- make cacao-native back to provide virtual/java-native,
and ecj-bootstrap-native back to provide virtual/javac-native
- Add back provider settings in README.
- Remove the update-alternatives in openjdk-8-native,
it doesn't work for native package.
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
|
|
This reverts commit 04d5d0bf414c05ca59618d77f17ff9898aa1c566.
Detail reason is in the following commit.
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
|
|
This adds openjdk-8 for native and target builds and allows a stripped
openjre-8 being built as well instead of trying to cherry-pick jre
components from jdk-image.
The recipes allow building openjdk-8 with or without:
* x11
* cups
* alsa/pulseaudio
and let packager enable unlimited-crypto, if desired.
To support certificate based java libraries, cacerts is created based on
ca-certificates from OE-core.
Since there can be only one PROVIDES for virtual/java-native and virtual/javac-native,
move the provides to openjdk-8-native (I think everyone agrees it's a better
choice than ecj-bootstrap-native).
Plus: Applying a fix from openjdk-9 repository which fixes build issues using gcc5
Signed-off-by: Jens Rehsack <sno@netbsd.org>
Signed-off-by: Maxin B. John <maxin.john@intel.com>
|
|
Instead of potential circular depending virtual/javac-native (even this
recipe provides such a useable java-native), rely on well known path
via cacao-native to build up to icedtea7-native in reliable manner.
virtual/javac-native should be used by parts not belonging to the bootstrap
phase.
Signed-off-by: Jens Rehsack <sno@netbsd.org>
Signed-off-by: Maxin B. John <maxin.john@intel.com>
|
|
Cacao-initial-native suffers from sporadic segmentation faults, when the Java
program uses multiple Java threads, due to a race condition. In particular,
jamvm-native and classpath-native builds would occasionally fail as the
eclipse compiler used during build task by default uses multiple threads.
As fixing the race condition would require large changes to the
cacao-initial-native codebase, and as it is only used briefly during the
bootstrapping process, it is easier and safer to just force ecj-initial to run
in a singlethreaded mode.
Signed-off-by: Erkka Kääriä <erkka.kaaria@intel.com>
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
|
|
Since bd9c6023, virtual/java-initial is provided by
cacao-initial-native and jamvm-initial-native. For consistency, this
commit makes them provide virtual/java-initial-native instead.
Signed-off-by: Mario Domenech Goulart <mario@ossystems.com.br>
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
|
|
A number of recipes in meta-java use autotools.bbclass for building; internally
it uses the sstate-control populate_sysroot manifests to locate and copy
aslocals for configure dependencies as part of the do_configure step.
The manifest path differs depending on if it's a native package or not so
autotools.bbclass looks at the package name to determine if its native or not
(it's native if it ends with "-native").
The warnings are emitted because a few (native) recipes in meta-java (such as
cacao-initial, classpath-initial and more) are incorrectly being classified as
non-native by autotools.bbclass, which makes it look for the manifest in the
wrong place and finally gives up with the warning:
WARNING: /home/oe/tmp-glibc/sstate-control/manifest-mymachine-libecj-bootstrap.populate_sysroot not found
This happens for cacao-initial, classpath-initial, ecj-initial, libecj-boostrap,
jamvm-initial and jikes-initial since they are all native and rely on autotools.
This patch renames the recipes so they end with -native, e.g ecj-initial-native.
It also updates the recipes so they DEPEND on the corresponding new name. This
helps autotools.bbclass to properly classify the recipes are native and silence
the warnings.
Signed-off-by: Magnus Olsson <magnus@minimum.se>
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
|
|
Conflicts:
recipes-core/cacao/cacao_1.6.1.bb
|
|
ecj recipe is built as initial and as native,
as such, unpackpost is run multiple times over
the same code source. Added check if a move
operation hasn't been already executed.
Signed-off-by: Cristian Iorga <cristian.iorga@intel.com>
|
|
|
|
When rebuilding in already built tree we have these
issues where a file is already deleted first time but
rm fails on reexecution. This patch adds checks for
such cases so that rebuilds are guarnteed
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
|
|
|
|
|
|
* on some filesystems find orders the files to build not the
right way, using sort seems to fix it, thanks Stefan Ring
for the patch
* use INC_PR for better revision handling
Signed-off-by: Henning Heinold <heinold@inf.fu-berlin.de>
|
|
Signed-off-by: Xerxes Rånby <xerxes@zafena.se>
|
|
* taken over mostly stuff from oe classic
* cleaned up recipes
* added license checksums
* bump icedtea6-native to 1.8.11
* use jamvm from git as native
|
Wes Lindauer
Douglas Royds
André Draszik
Otavio Salvador
Adrian Calianu
Patrick Ohly
Erkka Kääriä
Jackie Huang
Jens Rehsack
Mario Domenech Goulart
Magnus Olsson
Lukas Elsner
Cristian Iorga
Khem Raj
Henning Heinold
Xerxes Rånby