| Age | Commit message (Collapse) | Author |
|---|
|
bitbake/lib/bb/build.py:
bitbake/lib/bb/utils.py:
bitbake/lib/bb/shell.py:
bitbake/lib/bb/providers.py:
bitbake/lib/bb/msg.py:
bitbake/bin/bitbake:
bitbake/bin/bitdoc:
bitbake/classes/base.bbclass:
Start an overhaul of the message handling in bitbake:
- Introduce a new msg module to replace the existing simple calls.
- The msg module adds the conncept of message domains so ultimately we can
select which kinds of debug messages we want to see (it uses an Enum class
for this)
- Add a warn logging level for things the user should really pay attention to
as note is a little overloaded at present
- Start converting to use the new fuctions
|
|
bitbake/lib/bb/__init__.py:
bitbake/lib/bb/shell.py:
bitbake/bin/bitbake:
* Move providers functions into their own file. The arguments are horrible
at present but we have to start somewhere with the cleanup and refactoring.
|
|
Really use myFilterProvider which will take care
of the PREFERRED_PROVIDER, and add documentation
to the code.
|
|
Add code to handle the RDEPENDS. We use getRunProviders
and the rundeps variable of BBParsingStatus to get the
providers, similiar to addRunDeps.
The graph is now not looking too bad.
|
|
Due the recursion we need to reevualate seen_depends everytime.
Move the variables/files down to the actual call...
|
|
First version of showing DEPENDS and RDEPENDS in the same graph.
The is one issue that some provider get handled twice and we
see two arrows.
|
|
RDEPENDS needs different handling than DEPENDS. Looking
in providers is not enough most of the times.
|
|
|
|
Emit the Version of the provider in the node as well. Create
a myFilterProvider to take PREFERRED_PROVIDER into account.
|
|
If your dependency graph should not show specific
depends, e.g. the virtual/libc, quilt-native depends
use -I virtual/libc -I quilt-native to do so.
|
|
The code had the assumption that at least one .inc
file was included (always included). This is not true
in the NSLu2 case. The code now makes sure we load each
bbclass we have in the INHERIT line either directly or
indirectly. And then we can safely add all event handlers.
Todo: Add HANDLERS += to only load the HANDLERS and not
all classes.
|
|
Add option -g to generate graphs for
DEPENDS and RDEPENDS. Currently these
are emitted in two different files for now.
For the future:
-Add option to stop at level X
-Emit both DEPENDS and RDEPENDS into
the same graph...
|
|
BBConfiguration do not initialise the
configuration twice
Save a copy of th configuration data and update
it.
|
|
- Make the update_data for buildDepGraph act on a copy of self.configuration.data
fixing certain dependency errors.
|
|
This is a unstable BitBake version. Warn the
users about it... or annoy them enough to switch
to the stable branch.
|
|
-Make trunk 1.5.0 which will go to 1.6.0
-The stable branch is now 1.4.1 which will
be 1.4.2 on release
|
|
Merge -453 to HEAD into the trunk from the branch
|
|
Add hack to register even handlers seen by the configuration
file. We assume that we included at least one .inc file so
INHERITS += got used.
We should add a HANDLERS variable which will be responsible
for installing the handlers
|
|
* Fix a loadDataFull call the last changeset missed
* Only follow all dependencies when cmd == build
(fixes the recursive cleaning bug)
|
|
-Change version to 1.3.9 we will change it to 1.4.0 for the
final release
bitbake/lib/__init__.py:
-Change the version as well
|
|
bitbake/lib/bb/shell.py:
bitbake/bin/bitbake:
* Remove a useless return value from loadDataFull
bitbake/ChangeLog:
* Update the changelog with recent changes
|
|
* Detect and remove deleted files from the cache
* Add/enable/fix cache debugging
bitbake/bin/bitbake:
* Cache rdepends and rrecommends against filenames as well as pn
|
|
bitbake/lib/bb/__init__.py:
bitbake/lib/bb/data_smart.py:
bitbake/lib/bb/cache.py:
bitbake/lib/bb/shell.py:
bitbake/bin/bitbake:
* Major cache refactoring. Change the cache to store only the data bitbake needs
for dependency calculations instead of all the metadata.
* Separate the cache code into its own file.
* Update the rest of the code to work with the cache changes.
* Temporarily break the shell's poke command.
|
|
Update output to more accurately describe what bitbake's doing
when handling already staged packages.
|
|
-Do not add handlers again. It is 'fast' enough
to iterate over them in the BBHandler do not
add them in bitbake again.
This temporarily breaks the cache - which
will be fixed shortly
lib/bb/event.py:
-Have a data structure that operates faster on
lookups
-Change register/remove to take a 'name' parameter
lib/bb/parse/parse_py/BBHandler.py:
-Update to the new event.register syntax
|
|
Do not build the World Package List unless we want
to build 'world'. This saves from 10 to 20 seconds
on a usual build.
I think people doing a world build can survice the
extra seconds.
|
|
Fix RRECOMMENDS_${PN} and RPROVIDES_${PN} handling.
|
|
Create a ordinary main method. This makes invoking
the hotshot profiling tool much more easy
|
|
Make it work with python2.3 and python2.4
|
|
-Use paragraphs instead of 'pre'. This enables line
wrapping for too big comments
|
|
-Do not link to keys.html and groups.html, use the right
site
-Make it possible to have descriptions for groups as well
|
|
-Work on case insensitive filesystems where keys.html == keyS.html
-Create all_groups.html and all_keys site
-Sort the keys in a group
-Emit the ',' only when needed
|
|
-We do not need make, we only need bb.parse
|
|
Introduce a NoProvider Event. It contains
a 'runtime' flag, the item name and the
configuration data.
bitbake/bin/bitbake:
Fire the NoProvider event at some spots. There
are more events we need to fire. Ideally the
BitBake UI would just be a handler...
|
|
Prepend the lib module instead of appending to
override the system installation
bitbake/lib/bb/data.py:
Same as above, prepend instead of append
|
|
Remove the whitespace... well I have seen it and
removed it.
|
|
-make multiple providers inherit
Event directly. Save the kind (runtime vs. buildtime)
the item and the providers
bin/bitbake:
-Fire multiple providers issue as Event
|
|
Safe the number of failed packages even if the BuildCompleted
bin/bitbake:
count the number of broken packages
|
|
- ASSUME_PROVIDED allowed to include runtime names
|
|
Need to check both RDEPENDS and RDEPENDS_${PN} for RDEPENDS
|
|
Bump the version and hopefull we will end up as 1.3.4
|
|
Bump the version to 1.3.3
lib/bb/__init__.py:
Bump the version to 1.3.2.2
|
|
Stop hardcoding the stamps directory directly but
use the STAMP variable instead. The Package Version, Name
and Release is already included in the STAMP variable.
This patch was created by Phillip Zabel (pH5). Thank you
|
|
It could be that we have no buildable providers and
eligible is None. Do not error in this case.
This was spotted by mickeyl on a world build (with a
lot of failures) and fixed by Richard Purdie again.
|
|
Patch by Richard Purdie to fix the selection of
a preferred R provider
|
|
Patch by Richard Purdie to fix the recognition
of circular dependencies.
|
|
-add explode_deps method to return
a list of package names from a string (RDEPENDS)
Patch courtsey Richard Purdie (rpurdie@openedhand.com)
bitbake/bin/bitbake:
-major overhaul on dependency tracking and improvement
to the previous commit.
Build Runtime Depends and Runtime Recommendations at the
same time as the package is built.
Patch courtsey Richard Purdie (rpurdie@openedhand.com)
|
|
Build a reverse hash for PACKAGES, so runtimes dependencies
can be resolved.
E.g. which bb file is needed to get glibc-locale-de build.
Patch by Richard Prudie of zaurus kernel and openedhand.com
fame. Thanks!
|
|
patch by pHilipp Zabel on #531. The variable
was initialisied with one name and used with other
names.
I don't know if this lead to any exceptions but
it is better this way.
|
|
Patch for #436 to fix
CVSDATE = "xxx"
CVSDATE_somepackage = "yyy"
PREFERRED_VERSION_somepackage = "zzz-${CVSDATE}"
results in "zzz-xxx", not the expected and preferred "zzz-yyy".
This is due to PN not being in OVERRIDES at this stage of bitbake operation.
applied RP's patch from http://bugs.openembedded.org/attachment.cgi?id=365&action=view
thanks RP
|
Richard Purdie
Holger Hans Peter Freyther