[DO NOT MERGE] Repin #88
Conversation
|
I think this would finish building if we split off qtwebkit |
|
windows doesn't have space left on disk |
I don't know if that's advisable. I mean, it's better to talk to Microsoft first to get more time and space for the builds to finish. |
recipe/conda_build_config.yaml
Outdated
| MACOSX_DEPLOYMENT_TARGET: # [osx] | ||
| - '10.12' # [osx] | ||
| CONDA_BUILD_SYSROOT: # [osx] | ||
| - /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.12.sdk # [osx] |
There was a problem hiding this comment.
Remove the 2 entries above and do a rerender. conda-forge-ci-setup will add the correct path.
recipe/conda_build_config.yaml
Outdated
| @@ -0,0 +1,4 @@ | |||
| MACOSX_DEPLOYMENT_TARGET: # [osx] | |||
| - '10.12' # [osx] | |||
There was a problem hiding this comment.
Why is this 10.12, but in the recipe 10.10 is used in lots of places?
recipe/build.sh
Outdated
| export CXX=${GXX} | ||
|
|
||
| mkdir -p "${SRC_DIR}/openssl_hack" | ||
| conda install -c https://repo.continuum.io/pkgs/main \ |
|
@isuruf thanks for taking a look at this one! We need all the help we can get to bring this one home. @marcelotrevisani you have some experience building qt, if you can spare some minutes to take a look at this it would help us a lot. Right now we are experiment segfaults downstream with pyqt (conda-forge/pyqt-feedstock#48), and the openssl migration has stalled here with qt. |
| matrix: | ||
| win_cxx_compilervs2015vc14: | ||
| CONFIG: win_cxx_compilervs2015vc14 | ||
| CONDA_BLD_PATH: D:\\bld\\ |
There was a problem hiding this comment.
Can we change this to C:? When building qgis this seemed to give more drive space (the lack of which seems to be the reason why the build is currently failing). I'm not sure if there are any issues with doing this though...
|
Trying to build the recipe in the docker container (linux-anvil-comp7) I get this error: |
|
@looooo, did you get that error while running the recipe's tests? |
yes, the compilation went fine but error occured during the tests. Package is uploaded to this channel: |
Those are fantastic news!! Let me fix the tests problem right away. |
|
@conda-forge-admin, please rerender |
…da-forge-pinning 2019.04.18
|
@looooo, please try again. Things should work now. |
|
@ccordoba12 done, worked. Tests are passing. |
|
@conda-forge-admin, please rerender |
…da-forge-pinning 2019.04.18
recipe/build.sh
Outdated
| @@ -59,7 +59,7 @@ if [[ ${HOST} =~ .*linux.* ]]; then | |||
| export CXX=${GXX} | |||
|
|
|||
| mkdir -p "${SRC_DIR}/openssl_hack" | |||
| conda install -c https://repo.continuum.io/pkgs/main \ | |||
There was a problem hiding this comment.
I thought that conda-build did not even allow for that :-/
recipe/conda_build_config.yaml
Outdated
| @@ -1,4 +1,2 @@ | |||
| MACOSX_DEPLOYMENT_TARGET: # [osx] | |||
| - '10.12' # [osx] | |||
There was a problem hiding this comment.
@isuruf and @mingwandroid is the failure here related to 10.10? Should we force 10.12 for this build?
There was a problem hiding this comment.
I believe macOS SDK 10.10 is what AD targets. We don't have a conda_build_config.yaml for Qt (or at least didn't for 5.9).
There was a problem hiding this comment.
Thanks. I'm quite lost in all the commit here. I'm also trying from a "fresh" PR based on AR but there I'm getting many undefined reference to dbus stuff, like:
lib/libQt5WebEngineCore.so: undefined reference to `dbus_message_set_serial@LIBDBUS_1_3'
In the past we had dbus from yum_requirements.txt but now we have a package in host. Not sure how to fix that.
There was a problem hiding this comment.
I already fixed that in
There was a problem hiding this comment.
I'm fixing run_exports in dbus so we can try to keep the recipe as close as possible to AR.
|
Most of this PR is in Done in #102. Thanks to all 11 people who contributed to it!! |
Closes #87