Is the scylladb build hardcoded to work only with gnu gcc?

Kenneth 0
Viewed 35

The python configure.py contains a line

gcc_linker_output = subprocess.check_output(['gcc', '-###', '/dev/null', '-o', 't'], stderr=subprocess.STDOUT).decode('utf-8')

The comments before this line indicate scylladb uses a custom dynamic linker and references details about the ABI layout.

Is there code missing from the configure.py script which would enable building on a strict llvm environment, or is that not possible at this time?

I am building Scylladb on FreeBSD 13 which uses clang++ 13.0.0. I am on branch master, commit 0efdc45d5981868b1b6, Setp 8, 2022. I patched SCYLLA-VERSION-GEN to get past the date --utf and USAGE issues, and patched config.py with an entry to read ID from freebsd for the boost error message. I run configure.py with

./configure.py --mode=release --compiler=clang++ --cflags=-I/usr/local/include
scylla freebsd clang++ build
1 Answers

In fact ScyllaDB builds with clang. However its dependency Seastar is very dependent on Linux. If you want it to run on FreeBSD you'll have to port Seastar first (see reactor_backend.{cc,hh})

Avi Kivity0
Related