On Wed, 4 Mar 2026 22:11:34 GMT, Roger Riggs <[email protected]> wrote:
> I've gotten similar failures on Ubuntu 25-10, with a ENOENT (2) errno. Its > hard to identify which `goto WhyCan'tJohnnyExec` is the cause. (I don't have > an Ubuntu 25-10 yet to test with). Both the InheritIOTest and Basic.java (all > three fork modes) have failed looking like the exec fails. This is highly strange. And I got similar reports from AIX. I would love to reproduce this, but cannot reproduce it on Debian 13, Fedora 43, nor inside an ubuntu 25.10 container on Fedora 43. If it were only POSIX_SPAWN mode, my current suspicion would be that the kernel closes CLOEXEC file descriptors before dup2ing them in posix_spawn, like on MacOS. It could be related to the kernel version, which may explain why I cannot reproduce it in an ubuntu 25.10 container. But the really strange thing is that you report this error message (`Bad code from spawn helper`) from other modes too. This code should not even run on FORK mode. Do all three modes really fail with this exact error message? Does the error happen intermittent or always? Can you share the precise configure line and the jtreg invocation line that ran this test? ------------- PR Comment: https://git.openjdk.org/jdk/pull/29939#issuecomment-4003493890
