Community
Participate
Working Groups
Currently we build e4 0.11, and then run a "short" 4.1 SDK build that reconsumes these bundles and the 3.x bundles. We should instead run the 4.1 SDK build first and then later build any additional e4 bundles that are not in the SDK. Some reasons for this: 1) Source drops for legal review, currently a source drop for legal review would need to be assembled manually because the source compiled for 4.1 is split over two builds. 2) Problems in e4 bundles can block the 4.1 SDK build. If 4.1 was built first then problems in other e4 bundles would not affect it. 3) Chaining builds in Hudson, this would be the proper dependency order if we reach the point where this can all be chained together in hudson. 4) Testing, current e4 tests all run in a throw-away 3.7 + e4 install. The tests are probably better run against the 4.1 SDK
I have released some changes to org.eclipse.e4.sdk on a R4_1_SDK_LONGBUILD branch. This runs first and works when tested locally on my machine. We need a short version of the 0.11 e4 build and modifications to the bash scripts.
Created attachment 187227 [details] patch Here is the patch merging my changes from the R4_1_SDK_LONGBUILD branch back to HEAD. I will be releasing this to use in tonight's IBuild Some things to note: 1) The map files used to build the 4.1 SDK are the combination of org.eclipse.e4.sdk/maps + releng/*/maps for all the other projects 2) The map files used to build the e4 bundles after the SDK are the releng/*/maps as before. These maps are filtered so that any bundle that was build as part of the SDK is not rebuilt again. 3) Tests are now run against a proper 4.1 SDK install.
Builds look ok
