Re: Mangement of fdtable[]

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On Oct 24,  7:16pm, Peter Stephenson wrote:
} Subject: Re: Mangement of fdtable[]
}
} Is this the fix for tcp.c (not socket.c)?

This looks very similar to what I was contemplating, except that I had
not yet thought as far as adding another FDT_* flag.  The only other
thing I was wondering about is allowing the module to control whether
the FD is closed when entering a subshell [i.e., when exec.c calls
closem(FDT_INTERNAL)], because there is not always an execv() in that
case so simply setting CLOEXEC in the module is not enough.

} I didn't think FDT_INTERNAL was the right thing, on balance,
} as they're too public for the shell taking over management

That's true for tcp.c, but perhaps not so for e.g. db_gdbm.c.