| Age | Commit message (Collapse) | Author | Files |
|---|
|
stuff packaged.
Without this change opie_do_install was called after packaging.
Changeset is a result of moving do_stage after do_install
|
|
packages/quilt/quilt-native.inc: Run make check for quilt native
Run make check for quilt-native from the do_test task. This
will run the first native test. If the test fails quilt will
fail.
The tests will only be run if you do INHERIT+=test
|
|
inside ${D}
|
|
the task reordering
|
|
people are reporting
~lart kergoth for introducing this regression
|
|
* Add package "depchains". This facilitates, for example, ensuring that if
A depends upon B, then A-dev will RRECOMMENDS B-dev, and the same for the
-dbg packages.
|
|
other packages.
|
|
trunk), updated preferred providers in nylon.conf.
|
|
|
|
* Avoid premature use of the PKG_* variables. We don't need to make use of
the debian.bbclass (or otherwise) renamed package names until the very end
of the packaging. This was necessary in order to enhance my
depchain/correspondantdeps stuff, and doesn't seem to harm anything.
|
|
* This bbclass does path rewriting within *-config scripts for packages
being staged. However, it used to expect variable assignments in form foo=bar,
whereas some packages (e.g. libpng) have foo="bar". Fixed to understand both
formats.
* Note: I consider this fix a workaround, as it uses overquoted regexps passed
via command line. Instead, whole rewriting should be reimplemented in Python.
|
|
and '5db3d555762b1c8d96106bfae60bec656db47eba'
|
|
suffixed with version properly.
* Sometimes soname of a shlib is overriden to contain extra version. If
it's build by libtool, .la still has basename w/o suffix. This fix recognizes
such case, and handles shlib installation properly (via libtool).
* Example of this is smpeg_0.4.4+0.4.5cvs20040311 which builds libsmpeg-0.4,
while still being linked as -lsmpeg.
Fixes #1389.
|
|
and 'c2682ec1dd418a696aa5cec3d278f7cfd4386443'
|
|
* Fix bug which caused non-detection.
* Make visible error message for such condition.
* But still recover and continue for now, while we don't
have all occurances fixed in metadata. Note that while
content of the package will be ok, metadata can be wrong.
So, recover behavior should not be relied upon, this going to
be fatal condition later.
Oked-by: RP, hrw
|
|
with the funy BitBake scope
The def methods are in a different scope that the PACKAGEFUNC and the
result is only the packagefunc is able to use 'd' so we need to make
this a parameter for the other methods.
Use the installed scanelf to check for the RPATH of the files
|
|
and 'dbb377d00572c21544539346dc158feea2cfd1d6'
|
|
* Factor out "strippedness" substring as FILE_UNSTRIPPED_MATCH.
* Allow FILE_UNSTRIPPED_MATCH to be overriden, useful to support
other executable formats.
* Also, don't complain if .debug directory already exist
(can heppen if manually run BB tasks, e.g. for debugging).
|
|
single charachter
[2] will get us the 2nd charachter of the first line. This
is not what we wanted. We want the version number which is the
'2nd' column of the first line.
|
|
|
|
|
|
Make legitimize_package_name also convert <U0123> style encoding of
unicode codepoints into their utf-8 representation, as in glibc locale
files.
|
|
|
|
imported -before- the current patch rather than -after-.
|
|
doing a --dry-run, never actually applying the patch. Only quilt-native in oe was using that.
|
|
|
|
* switch os.mkdir to os.makedirs.
* pass on all errors from QuiltTree.Clean(), as it can fail in ways that do
not need to be reported to the user, and a failure will end up being seen
again during the Import/Push of the patches.
|
|
* Add NOOPResolver class, which simply passes the patch failure on up, not
doing any actual patch resolution. Set PATCHRESOLVE = "noop" to make use of
it. Most useful for unattended builds.
|
|
the quilt patcher class. Without this quilt will search for a patches
directory - starting from the current directory up to the root
directory. If it finds an existing patches directory it will use it
for its patches. This causes all sorts of problems since it is not
where the patches are expected to be. Prior to the recent patcher
changes this directory was being created.
|
|
Rework the way patches are handled. There are now two abstract base classes,
initialized in patch.bbclass. One for patchset operations on a directory, and
another for patch failure resolution. Currently includes 'patch' and 'quilt'
concrete PatchSet classes, and a 'user' resolver class, which simply drops you
into a shell in the source tree to fix the rejects.
|
|
|
|
|
|
for the new task-machine code (from poky)
|
|
|
|
|
|
to. Should be backwards compatible.
|
|
package if we want
|
|
|
|
* this will force a full rebuild, so have some coffee + cookies ready
|
|
and 3608f3d5434271fde9ce2b2c787339675d8a532f
|
|
multithreading version of bitbake. Also, set BB_DEFAULT_TASK to specify the default task (build) rather than hardcode into bitbake.
|
|
|
|
For icecream we need to package a tar.bz2 containing our
used toolchain. It must contain a gcc, g++ and cc1 and cc1plus
and a spec file (if it is not built-in). Optionally it contains
host libc.so and libfd.so (which we do not package). Simplify the
code creating the tar.bz2 and make it work with any version of
gcc/glibc/uclibc (untested).
|
|
os.path.join to create paths
|
|
native or cross package
|
|
As requested by Marcin include the PR of the built package as well. This
is already included in the PF variable.
|
|
|
|
|
|
Fixup truncate to forget the the status of previous runs. We will
simply write("") to the file and empty it this way. int(f.read()) will
fail then
|
|
Using the continue option of BitBake can lead to some issues with the
Tinderbox. Even if one build failed the color of the result will be green
as the BuildCompleted event will be send by BitBake.
As a workaround we will store the failure in a tinder-status file in case
of failure (e.g. no Provider or PkgFailed) and will reread on the start.
When the build starts we will truncate the file and remove 'the' error
|
