time usage for subprocesses [NEW PATCH]

[Zack Weinberg]

Jeffrey A Law wrote:
>   In message < 9908131132.AA20523@vlsi1.ultra.nyu.edu >you write:
>   >     I don't see what the problem is.  It is not like this is a critical
>   >     feature, or a feature likely to be used at all except by maintainers
>   >     or the curious.  If you are on a broken system, don't use -pipe (it i
> s
>   >     sort of like the old joke, Docter, docter, it hurts when I do this --
>   >     don't do that).
>   > 
>   > If a feature won't work, we shouldn't document it.  Who knows who will
>   > use it?  All sorts of features get used in ways we don't expect or by
>   > people we don't expect to use them.
> Or we could document it as "this features does not work if you do XYZ" or
> "this feature only works if your host system has feature PDQ".  Or have the
> compiler turn off -pipe when using this option.

I sent a third version of the patch which turns off -pipe when -time
is given.  Someone said they would like to allow the combination when
we have wait4, which is a fine thing to want except that it involves
resurrecting "pwaitr" which I didn't like very much.  Furthermore, it
turns out that some systems (*cough* Solaris) fake wait4 with
getrusage and therefore you'll get garbage numbers if you have two
processes executing at once.  All in all I don't think it's so bad
that -pipe -time doesn't work.

zw