Alkis Evlogimenos
Alkis Evlogimenos
Friendly ping. @emkornfield @pitrou could you take a look?
> Looks like CI failure might be related to this change: > > ``` > FAILED: debug/parquet-dump-footer > : && /usr/lib/ccache/c++ -Wredundant-move -Wno-noexcept-type -fdiagnostics-color=always -Wall -Wno-conversion -Wno-sign-conversion -Wdate-time -Wunused-result -fno-semantic-interposition...
> I think that the build doesn't have `-DARROW_BUILD_STATIC=ON`. > > BTW, we should not use `${ARROW_LIBRARIES}`. That is `arrow_shared`, `arrow_static` or `arrow_shared;arrow_static`. If it's `arrow_shared;arrow_static`, both of them are...
> Could you try reverting the commit? There are link errors now on windows and macos: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/50227715. Any help on cmake would be greatly appreciated.
> Could you try linking `parquet_shared`/`parquet_static` instead of `arrow_shared`/`arrow_static`? `parquet_shared`/`parquet_static` are linked for all tools a few lines above: https://github.com/apache/arrow/pull/42174/files#diff-94a0cc5cd65c55d279af9f4d93dd7d45cb2ad1bb4b12c463c3a3023e32b6c94dR24-R28 so I removed the `array_shared`/`arrow_static` stanza.
> > Could you try linking `parquet_shared`/`parquet_static` instead of `arrow_shared`/`arrow_static`? > > `parquet_shared`/`parquet_static` are linked for all tools a few lines above: https://github.com/apache/arrow/pull/42174/files#diff-94a0cc5cd65c55d279af9f4d93dd7d45cb2ad1bb4b12c463c3a3023e32b6c94dR24-R28 so I removed the `array_shared`/`arrow_static` stanza. Still...
> Hmm. It seems that auto generated classes from `parquet.thrift` such as `parquet::format:SizeStatistics` don't have `__declspec(dllexport)`/`__declspec(dllimport)` for Windows. So they can't be used out of `libparquet`. > > Do we...
> It will work. Could you try it? That worked! Thank you @kou.
> @alkis I think I might have misphrased something here or misdesigned something here. Where do you see the require second fetch as a requirement. I imagine performant readers will...
Perhaps the option should be `aggregates` instead of `no_constructor` and `class`es be made into `struct`s instead?