[Bug 228261] Request for a specific llvm commit to be merged into current and llvm60 port

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Tue May 15 03:09:48 UTC 2018


            Bug ID: 228261
           Summary: Request for a specific llvm commit to be merged into
                    current and llvm60 port
           Product: Base System
           Version: CURRENT
          Hardware: amd64
                OS: Any
            Status: New
          Severity: Affects Some People
          Priority: ---
         Component: bin
          Assignee: bugs at FreeBSD.org
          Reporter: FreeBSD at ShaneWare.Biz
                CC: brooks at FreeBSD.org

I am the port maintainer for devel/godot

The automated port builds for current have been emailing me for a while about a
build failure of the godot port. Turns out that one particular file takes over
1.5 hours to compile and is normally killed as a runaway. My poudriere builds
don't kill runaways and I hadn't noticed the 2 hour build times.

The issue is related to an optimisation, so should be a more general issue than
one ports code base. Using -O0 bypasses this issue but -O1 or higher hits it.
It is also amd64 specific, adding -m32 bypasses the issue.

This build issue has been noticed by other godot developers (on osx and
debian), and a commit has made to llvm-master that resolves it which has not
been merged into the 6.0 release. Using clang++-devel from
llvm-devel-7.0.d20180225 does not hit this issue.

The same patch should be applied to current as well as the llvm60 port.

The llvm commit that reportedly fixes this is

The godot project issue related to this is

You don't need to build the entire godot port to test this, you can start
building the port, then stop it after the first files compile and just compile
the one file, if it takes more than 10 seconds you have hit the issue.

cd /usr/ports/devel/godot
cd work/godot-3.0.2-stable
clang++60 -o test.o -O3 -I. -Icore -Icore/math -Iplatform/x11

You are receiving this mail because:
You are the assignee for the bug.

More information about the freebsd-bugs mailing list