This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: system and popen fail in case of big application
On Sun, Sep 9, 2018 at 10:47 PM Martin Buchholz <martinrb@google.com> wrote:
>
> See discussion in the source file
> http://hg.openjdk.java.net/jdk/jdk/file/tip/src/java.base/unix/native/libjava/ProcessImpl_md.c
Thanks, that's useful information on the bigger picture.
FYI, the test program in PR #10311 runs to completion without error
under glibc 2.27, when compiled on x86_64-linux with either -m32 or
-m64. I think this probably reflects our backing away from caching
the result of getpid() a few versions ago. I don't want to say that
this has been _fixed_, but if you gave us a list of everything you
would actually want to do on the child side of a CLONE_VM without
CLONE_THREAD, I think we could now have a more constructive
conversation about whether that's something we could commit to keeping
working.
It's interesting that that cites not being able to reliably close all
inherited file descriptors as a reason not to use posix_spawn. I have
thought for years that closefrom() should be universal, and I have yet
to hear an argument against it that wasn't, in my opinion, 100% bogus.
Perhaps we should push for it to be adopted in Linux.
zw