refactor(c)!: Updated include/install paths for adbc.h #1965
Conversation
c/CMakeLists.txt
Outdated
| @@ -35,6 +35,8 @@ add_subdirectory(vendor/nanoarrow) | |||
| add_subdirectory(driver/common) | |||
| add_subdirectory(driver/framework) | |||
|
|
|||
| install(FILES "${REPOSITORY_ROOT}/c/include/adbc.h" DESTINATION include/arrow-adbc) | |||
There was a problem hiding this comment.
I might be misunderstanding when we want this to be installed, but in main the driver_manager CMake configuration is responsible for installing the adbc.h file; figured it made more sense in the top level config
lidavidm
changed the title
.github/workflows/native-unix.yml
Outdated
There was a problem hiding this comment.
I think we still want to trigger workflows when the header is updated?
There was a problem hiding this comment.
Ah, it's under c/include now, cool.
c/include/adbc.h
Outdated
There was a problem hiding this comment.
Perhaps we can do this in two parts but I'd also like our own sources to use the 'proper' arrow-adbc/adbc.h include path
There was a problem hiding this comment.
Sounds good - I think probably worth doing in two parts. From what I see so far there might need to be a few things shifted around with all of the other languages
|
Not sure I understand why the R build for Windows is failing - hoping @paleolimbot may know |
#1966 should do it! |
WillAyd
force-pushed
the
fix-adbc-include
branch
2 times, most recently
from
377f130
to
5cf439f
Compare
|
Thanks. Could you try the following? diff --git a/ci/linux-packages/debian/libadbc-driver-manager-dev.install b/ci/linux-packages/debian/libadbc-driver-manager-dev.install
index 6cee53d04..933e26364 100644
--- a/ci/linux-packages/debian/libadbc-driver-manager-dev.install
+++ b/ci/linux-packages/debian/libadbc-driver-manager-dev.install
@@ -1,5 +1,4 @@
-usr/include/adbc.h
-usr/include/adbc_driver_manager.h
+usr/include/arrow-adbc/
usr/lib/*/cmake/AdbcDriverManager/
usr/lib/*/libadbc_driver_manager.a
usr/lib/*/libadbc_driver_manager.so
diff --git a/ci/linux-packages/yum/apache-arrow-adbc.spec.in b/ci/linux-packages/yum/apache-arrow-adbc.spec.in
index 9af14a2bf..70105cb06 100644
--- a/ci/linux-packages/yum/apache-arrow-adbc.spec.in
+++ b/ci/linux-packages/yum/apache-arrow-adbc.spec.in
@@ -120,8 +120,7 @@ Libraries and header files for ADBC driver manager.
%defattr(-,root,root,-)
%doc README.md
%license LICENSE.txt NOTICE.txt
-%{_includedir}/adbc.h
-%{_includedir}/adbc_driver_manager.h
+%{_includedir}/arrow-adbc/
%{_libdir}/cmake/AdbcDriverManager/
%{_libdir}/libadbc_driver_manager.a
%{_libdir}/libadbc_driver_manager.so |
|
BTW, should we install // adbc.h
#include "arrow-adbc/adbc.h"// adbc_driver_manager.h
#include "arrow-adbc/adbc_driver_manager.h" |
No strong preference on this so will leave up to @lidavidm |
Once we get things working here I think a compatibility header is a good idea (at least for a few versions) |
|
I think the C# failures are unrelated, so just down to R at the moment. Can't seem to reproduce locally so any guidance there is greatly appreciated |
|
Rebasing would make the C# failures go away, I think. |
|
@paleolimbot I know you've been super busy lately but if you get any time to advise on the R failures would be greatly appreciated |
|
Sorry for the delay circling back! There was a CI issue with |
WillAyd
force-pushed
the
fix-adbc-include
branch
from
ab40814
to
84549ce
Compare
|
OK should be green, and I've added the old headers in the standard include path with a pragma message to discourage their use. Let me know what else we may need on this |
There was a problem hiding this comment.
Thanks for working through this!
I'll leave this for @paleolimbot and @zeroshade in case they want to have a final look at the R/Go changes as well
…2051) The R check actions are currently failing because an update to r-lib actions resulted in some quarto actions being invoked, and these have not yet been whitelisted for use in Apache repositories. We don't actually need quarto for anything in the R package; however, the action is written in such a way that even though the step is skipped, the action is still loaded (and therefore its name is checked against the whitelist). Unrelatedly, the extended R check was failing because I forgot to copy a change from #1965 into one of the workflows that only runs once a week.
closes #1150