DuckDB 1.1 undefined reference to atomic

[vgrechin]

What happens?

I'm building DuckDB 1.1 on RaspberryPi 5 (Bookworm, 32-bit edition), with this

~/build/duckdb-1.1.0/build $ cmake .. --install-prefix /opt/duckdb
make -j4

Duckdb11Configuring.txt

It looks like libatomic isn't linked properly, so that encounter multiple errors of this kind:

parquet_extension.cpp:
(.text._ZN6duckdb19ParquetScanFunction21ParquetScanInitGlobalERNS_13ClientContextERNS_22TableFunctionInitInputE[_ZN6duckdb19ParquetScanFunction21ParquetScanInitGlobalERNS_13ClientContextERNS_22TableFunctionInitInputE]+0x750): undefined reference to `__atomic_store_8'
collect2: error: ld returned 1 exit status
make[2]: *** [tools/sqlite3_api_wrapper/CMakeFiles/test_sqlite3_api_wrapper.dir/build.make:130: tools/sqlite3_api_wrapper/test_sqlite3_api_wrapper] Error 1
make[1]: *** [CMakeFiles/Makefile2:7608: tools/sqlite3_api_wrapper/CMakeFiles/test_sqlite3_api_wrapper.dir/all] Error 2
make: *** [Makefile:136: all] Error 2

May someone take a look at the build scripts?

To Reproduce

~/build/duckdb-1.1.0/build $ cmake .. --install-prefix /opt/duckdb
make -j4

OS:

Raspberry Py OS Bookworm, 32 bit

DuckDB Version:

1.1

DuckDB Client:

CLI

Hardware:

No response

Full Name:

Vyacheslav Grechin

Affiliation:

Freelance

What is the latest build you tested with? If possible, we recommend testing with the latest nightly build.

I have not tested with any build

Did you include all relevant data sets for reproducing the issue?

Yes

Did you include all code required to reproduce the issue?

• Yes, I have

Did you include all relevant configuration (e.g., CPU architecture, Python version, Linux distribution) to reproduce the issue?

• Yes, I have

[szarnyasg]

Hi @vgrechin, thanks for reporting this. We do not provide official support for 32-bit architectures but we welcome community contributions towards ensuring that DuckDB runs on these platforms.

[vgrechin]

It is very unfortunate the architecture is not officially supported, because Raspberry itself is a great educational platform.
Just before considering my contributions to DuckDB codebase, please note that adding the flag of
-DDUCKDB_EXTRA_LINK_FLAGS="-latomic" suffices to build the code.

It may be wise though, to build some extra extensions immediately. Because them unavailable for download from web.
One's resullting build configuration might be:

cmake .. --install-prefix /opt/duckdb \
    -DBUILD_EXTENSIONS=httpfs\;json\;parquet \
    -DDUCKDB_EXTRA_LINK_FLAGS="-latomic"
make -j4 install

I gave the build a try against several random tutorials happily, while others may also work potentially
https://www.timestored.com/data/duckdb/find-time-series-peaks
https://www.timestored.com/data/duckdb/moving-average

[szarnyasg]

Hi @vgrechin that's an excellent point. We'll discuss supporting this platform next week. For now I added build instructions to the documentation via duckdb/duckdb-web#3639

[vgrechin]

Thanks, I left my comment there
Personally I pretty happy about size of parquets which the build of DuckDb is cabable to process on just 32 bit board.
For example, such fhvhv_tripdata_2020-07.parquet from NYC TLC dataset - roughly 300MB, 10M records. Although larger ones problematic due to memory constraints.

[github-actions]

This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 30 days.