It turns out that the way I use Python to create ZIP
archives is plain wrong. Some files simply don't have the
right size and we get EOFException when unzipping with
Java's ZipFile.
This change adds a flag to instead simply copy all the
files we want and then run the 'zip' system command.
The python zip facility is left there intact in case
I want to go back and fix it later (I'm going to assume
it's my usage that's wrong before really blaming python).
Change-Id: Iea178a49be0bf23c91c01a2e036ae7a76def2b55
Also fix packaging of system image (it had an extra root
directory in it, e.g. it was packaging android-4/armeabi-v7a/
instead of just armeabu-v7a/)
Change-Id: I0254b2680168d25103222f8871b2af37a7406b58
Also:
- in the images_*_source.props, we can't have the platform
version name (the human readable string). It's not a valid
property of the XML.
- disable the included-abi in the platform.
Change-Id: I3db62fde5e436bbe8f8e69eb1495ca4e6b954ba2
Somehow the build server doesn't automatically find atree.
In a future CL this should be amended to pass the proper
OUT_HOST_EXECUTABLE, being careful that it points to the
host/linux dir, not the current Windows one.
Cherry-pick from master b2b5921
Change-Id: I13e532cdc4c0839b6fd7f2a3a149725940a02734
This moves the copy/strip/rm logic for the Windows SDK
from the patch_windows_sdk.sh to a regular sdk-windows-x86.atree.
The bash script is still here to invoke atree and do whatever
cleanup we might want to do later.
This should make it easier to maintain the Windows SDK
by providing a structure similar to tools.atree.
This change requires build.git Change-Id: I22aae7a8 to
have rm/strip support in atree.
Cherry-pick from master e07f59a
Change-Id: Ia71fc69717eaee6e4d8d062144347db3326e4127
When the "sdk_repo" modified is used during the SDK build,
this also generates the repository.xml and addon.xml files
for the new SDK repository packages and places them in
the DIST_DIR.
Example usage:
make -j 16 PRODUCT-sdk-sdk showcommands sdk_repo dist DIST_DIR=/dist/sdk
make -j 16 PRODUCT-google_sdk-sdk_addon showcommands sdk_repo dist DIST_DIR=/dist/addon
Also merges 3ea87ca65f
'Fix XMLNS detection in SDK Repo XSD files.'
to prevent a merge conflict later on.
Change-Id: Ic9832e71456d76f849406e641792f4285ffbf649
When the "sdk_repo" modified is used during the SDK build,
this also generates the repository.xml and addon.xml files
for the new SDK repository packages and places them in
the DIST_DIR.
Example usage:
make -j 16 PRODUCT-sdk-sdk showcommands sdk_repo dist DIST_DIR=/dist/sdk
make -j 16 PRODUCT-google_sdk-sdk_addon showcommands sdk_repo dist DIST_DIR=/dist/addon
Change-Id: Ibaad66f4b1cc476a4146afde7087a10a0557f74c
When the build is invoked with the fake target "sdk_repo" and
a main target of sdk, win_sdk or sdk_addon, we now create
packages in DIST_DIR that can directly be used to populate the
SDK Repository.
This is quite close to how we actually distribute the SDK.
Change-Id: I01c729eff4dbc1eccbc7c5b1869f329363f1ce07
This removes the test that makes the SDK only strip llvm-rs-cc.exe if present.
It means the SDK will FAIL if llvm-rs-cc wasn't build successfully for Windows.
Change-Id: I4c73954c30265e71e84916976f16ae76d67a8e13
Firs this fixes a cp/rm errors that were logged when building the Windows SDK.
There errors happens because the secondary sdk/tools/patch_windows_sdk.sh
was run twice.
Next the shell script is changed to make any error fatal and actually
break the build, like it should.
In the makefile, merge Ying's fix to correctly pass showcommands.
Change-Id: Iee75e42b0b0bbba3c26858c82c933660dc6d74ae
The main makefile+scripts to build the Windows SDK is still this
one here in development/build/tools. However it defers to a new
matching set of files in sdk/build to build and package
things that depend on the sdk.git or external/qemu.git.
This will make it easier for us to prepare SDKs based on
a tools_rN branch that isn't cut at the same time than the
platform branch.
This is a multi-part changeset. The other part is in sdk.git.
This change also definitely removes support for building
the Windows SDK under Cygwin. Only building a specific subset
of individual binaries is supported at this point.
Change-Id: I4e9a2d810cf29fae0097fbd92be0cef89c9b3505
Merge commit '66d61ef95f5a0f8abc8b5371409e3c5372c9e10a' into gingerbread-plus-aosp
* commit '66d61ef95f5a0f8abc8b5371409e3c5372c9e10a':
Add in MonkeyRunner to the Windows SDK.
That's because GB uses the Tools_r& which do not have
hierarchyviewer2 so we should not try to pull the .bat from
hv2. We don't care since we'll ignore these old tools anyway
when packaging the final SDK.
Change-Id: Ie7d161bad59d6e919613db2e854f48964e59c4a9
- Move adb to the platform-tools.
- Update the windows build for it.
- Split the tools component from sdk.atre into sdk/build/tools.atree
Change-Id: I75608f12d6e1d8bc0f470e3ca26c7e550cfdcbd6
Merge commit '4336dcc64ba518b7c8bd0e4a9a9814064eb8920c'
* commit '4336dcc64ba518b7c8bd0e4a9a9814064eb8920c':
Update win_sdk to put binaries in SDK/platform-tools.
This is requires as a follow up to Change I4526310a
with moved platform tools from $SDK/platform/*/tools
to $SDK/platform-tools/
SDK Bug: 2897406
Change-Id: I1563c88fc824db3d9c86fb65479820989d46895c
The makefile makes it possible to "make PRODUCT-sdk-win_sdk".
It builds the Linux SDK, mirrors it as a Windows SDK and then use the shell
script to patch in all the Windows binaries.
(Merge master Change Ie24f765b)
The makefile makes it possible to "make PRODUCT-sdk-win_sdk".
It builds the Linux SDK, mirrors it as a Windows SDK and then use the shell
script to patch in all the Windows binaries.
Change-Id: Ie24f765bbe5201266efcb1c4cdcdead4ced35ef6
SDK Bug 2478848: emulator fails to build because sdl-config
was not checkout as +x by git/cygwin.
SDK Bug 2479321: aapt built using Cygwin 1.7.x depends on some
new cygwin DLL instead of mgwz.dll. The latter DLL is not even
present and fails to be bundled. This does not change the behavior,
it just makes sure we are using the proper version of cygwin.
