On 2/28/2019 2:04 PM, Dmitri Maziuk via Bacula-users wrote:
On 2/28/2019 8:06 AM, Andrea Venturoli wrote:
On 2/28/19 2:23 PM, Martin Simmons wrote:
That suggests the function accept4 was defined at compile time by
fails with
errno=38 (ENOSYS) at run time.
I don't know what to answer.
I'm not competent enough to understand the difference between the
two; I cannot also tell why it is defined and why it does not work.
It looks like accept(4) is a gnu-ism that takes a couple of
convenience flags and falls back on accept() if the flags are 0.
Arguably it's an error between the programmer's chair and keyboard: by
using a non-standard extension that at best saves you a couple of
extra lines of code, they lost portability to non-glibc systems
including presumable musl-based linux distros.
Taking that programmer's chair and keyboard error a step further...
There are only two flags that can be used in the accept4() flags arg,
SOCK_NONBLOCK and SOCK_CLOEXEC, setting O_NONBLOCK and/or O_CLOEXEC.
AFAIK, the O_CLOEXEC is a Linux-ism introduced in 2.6.23 and so
automatically makes it less portable. The O_NONBLOCK simply saves a
fcntl() call, so why does accept4() even exist. IMO, the original
programmer's chair to keyboard error was made by Gnu. This is a nearly
pointless POSIX extension.
But hey, we don't need portable C: ./configure libtool will figure it
all out for us.
Definitely agree with your doubts about libtool.
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
