Thank you for your quick response! I just took a few hours and successfully set up and built from WSL 1. It indeed is even faster than cygwin. Meanwhile, it's fascinating that the relative directory problem doesn't exist on WSL 1, that passing build/jmh/jars allows configure to locate JMH just fine.
Since WSL 1 and WSL 2 have such dramatic differences, should we briefly mention in building documentation that building with WSL should be primarily done with WSL 1, given now the default WSL installed from Microsoft Store only supports WSL 2? Gratefully, Chen Liang On Mon, May 8, 2023 at 8:36 AM <erik.joels...@oracle.com> wrote: > > Hello Chen, > > On 5/5/23 22:42, - wrote: > > Hello, > > While I was trying to diagnose the problem that JMH benchmarks can't > > be run on a fresh cygwin or WSL environment, > > https://bugs.openjdk.org/browse/JDK-8305669 I found myself troubled > > with path issues and slow building speed with a Windows target. > > Meanwhile the linux target is totally fine. > > > > 1. When I pass --with-jmh=build/jmh/jars in WSL configuration, it > > always fails to find the relative directories (not present on linux > > configuration). C:\\...\\jdk\\build\\jmh\\jars doesn't work. Only > > /mnt/c/.../jdk/build/jmh/jars work. Is this a bug? > I would say it's a bug. > > > > 2. My Windows WSL build is extremely slow. My windows target has > > almost stalled that it stuck on creating benchmark jar without even > > progressing to indify. And it seems WSL (Ubuntu) itself is only using > > 400MB of RAM with almost no CPU usage. However, these problems don't > > happen for my Linux target builds; meanwhile my cygwin builds are > > blazingly fast. What could be the cause of this slowness? > > > Are you running WSL 1 or 2? From what I've heard, WSL 2 doesn't work > well when having to interact with native Windows tools (which the > windows build has to do). Most reports I've heard about WSL 1 say it's > slightly faster than Cygwin, but different people and different > environments have different results. > > /Erik > >
