ninja doesn’t work.

[ytrezq]

ninja is a build tool similar to make required to build chromium.

However, using it never work :

fatal: posix_spawn: Invalid argument

Windows build : 10.0.14393

[therealkenc]

Ninja proper has been working for me here since well before the anniversary update (I think?) so that's curious. If you post a link to a gist with the output of strace -ff ninja <whatever> that will help track down what's going on.

[skyem123]

I'm not sure if I should post this here or in alwsl/alwsl#33, as I'm using it to get Arch Linux on WSL, however hopefully this is helpful, as I've found out how to reproduce this.

I have the same issue, whenever ninja tries to run a program, it errors with fatal: posix_spawn: Invalid argument.

Here is an example of trying to build ninja itself:

[skye@SKYE-WINDESK-01 ninja]$ ./configure.py --bootstrap
bootstrapping ninja...
wrote build.ninja.
bootstrap complete.  rebuilding...
[0/27] INLINE build/browse_py.hninja: fatal: posix_spawn: Invalid argument
Traceback (most recent call last):
  File "./configure.py", line 682, in <module>
    subprocess.check_call(rebuild_args)
  File "/usr/lib/python3.5/subprocess.py", line 581, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['./ninja']' returned non-zero exit status 1

Here is an strace log (strace -ff ninja -o log.txt): https://gist.github.com/skyem123/cfb3197a303a69d2f3fabeefb222abad

Hopefully this is useful!

[therealkenc]

That's clone(child_stack=0x7f341a850ff0, flags=CLONE_VM|CLONE_VFORK|SIGCHLD), aka perennial #1005. You're on Arch though, so you could be looking at a more general problem at a lower level, like is being observed with Ubuntu Yakkety here . Ninja 1.7.1 still runs as advertised here on WSL Xenial anyway.

[skyem123]

I'm not too sure how useful this will be, but I downloaded a copy of ninja from https://github.com/ninja-build/ninja/releases, and tried to run that, and I still got the same issue.

[therealkenc]

Gave me an excuse to upgrade to 1.7.2. Yeah it works fine on Xenial with 14971.

[skyem123]

Hm... I just checked my Windows build and it's 14393.447
I guess that means that the it's probably already fixed and I just need to wait until there's an update for it? (I'm hesitant to try the Windows insider stuff, though)

[therealkenc]

No the EINVAL on CLONE_VM|CLONE_VFORK|SIGCHLD in the trace you posted is still broken in 14971. So you're in luck, you don't need to try Insiders ;).

[skyem123]

This is interesting, when I run Ninja now on the Creators Update (still on Arch not Ubuntu) it doesn't have this error, it has a different one! In this case it seems to be related to #1878, as it has a very similar error to that.

[skye@Skye-WinDesk-01 ninja]$ ./configure.py
wrote build.ninja.
[skye@Skye-WinDesk-01 ninja]$ ninja
[0/27] INLINE build/browse_py.hninja: ../sysdeps/unix/sysv/linux/spawni.c:360: __spawnix: Assertion `ec'>= 0' failed.
Aborted (core dumped)
[skye@Skye-WinDesk-01 ninja]$

Note that for this, ninja is installed from the package manager. Bootstrapping ninja has the same error, this just shows it clearer.

The output of strace -ff ninja is... (after getting a working version of strace! bleeding edge on non standard systems is fun (my friend built version 4.14 for me on a real linux system)! actually building strace gives me the same error from the m4 command...) https://gist.github.com/skyem123/cdcc08668731e41738ca43fc31e50675.

Hopefully this is helpful!

[aseering]

Hi @skyem123 -- thanks for posting the strace! Unfortunately, it looks like the -ff flag didn't take effect: ninja is trying to get a child process, but we didn't get separate lines for it.

That said, it appears to be failing immediately after the clone() call (to start the child process). I wonder if any clone() experts are lurking and recognize an issue here:

mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_STACK, -1, 0) = 0x7f617b570000
rt_sigprocmask(SIG_BLOCK, ~[], [HUP INT TERM], 8) = 0
clone(child_stack=0x7f617b570ff0, flags=CLONE_VM|CLONE_VFORK|SIGCHLD) = 29425
write(2, "ninja: ../sysdeps/unix/sysv/linu"..., 87) = 87

Also, for what it's worth, I use ninja from Ubuntu 16.04's default repositories (ninja v1.5.1) and I don't see this issue. As a workaround, maybe downgrading would help?

[therealkenc]

This is a dupe of #1878 (or vice-verse). Which is why unzipping the locales manually is pointless, because glibc-recent will just hit the path on other stuff like ninja, m4, probably clang, etc.

[therealkenc]

The OP on this one was:

fatal: posix_spawn: Invalid argument

There was no strace provided so we'll never know; but best evidence this was #1005, because ninja wasn't being run on the supported 16.04 Xenial with glibc 2.23, and later unrelated #1878 (same reason).

[ytrezq]

I am confirming ninja is working… But only in latest development buid.