[PATCH] [PORTABILITY] don't use nonstandard __fpending in nohup
Rich Felker
dalias at aerifal.cx
Tue May 23 02:39:23 UTC 2006
On Sat, May 20, 2006 at 02:31:47PM -0400, Rob Landley wrote:
> On Thursday 18 May 2006 8:56 pm, Rich Felker wrote:
> > this patch may be controversial but it both reduces size and allows
> > building without using the nonstandard glibc-internal __fpending
> > function. i'm not sure whether the semantics are correct but EBADF
> > should only happen under very limited conditions.
> >
> > rich
>
> Oh good grief.
>
> The big long rant about fclose_stdout... Where did that come from? The only
> filesystem so pathologically broken that flush() won't actually flush the
> data and close() does is nfs, and this is nohup we're talking about; writes
> to stdout from nohup might as well just be discarded. We don't check the
> result of close from _cat_. We DON'T CARE. If you care, you call flush().
> If you can't get reliable results from that your filesystem is broken. (You
> can't get nfs anyway, you can just get less unreliable ones, and jumping
> through hoops to work around their strange cacheing just isn't worth it...)
>
> Good grief, fd_reopen isn't even jumping through these hoops when it closes
> the handles its' whacking. Talk about unnecessary paranoia...
>
> Time to weed-whack nohup...
I second this totally. I just commented on the implications because it
looks like someone went to great lengths to put those hacks in place,
and I figure I shouldn't be presumptuous about what should be removed
or fixed even when it's a big inconvenience to me and non glibc users.
:)
Rich
More information about the busybox
mailing list