Mail Archives: cygwin/2002/06/26/12:12:19
--- Christopher Faylor <cgf AT redhat DOT com> wrote:
> On Tue, Jun 25, 2002 at 10:06:38PM -0700, Rick Rankin wrote:
> >--- Christopher Faylor <cgf AT redhat DOT com> wrote:
> >> On Tue, Jun 25, 2002 at 09:09:43PM -0700, Rick Rankin wrote:
> >> >Corinna mentioned earlier in this thread that some strace output might be
> >> >helpful. I've attached two strace logs from running 'ls' in an empty
> >> directory.
> >> >The first is using the 1.3.10-1 DLL where don't see the slowdown, the
> second
> >> is
> >> >using the 20020625 snapshot of 1.3.11 where I do.
> >>
> >> The microsecond numbers in the 1.3.10 vs. 1.3.11 strace show that 1.3.11
> is
> >> actually slightly faster than 1.3.10:
> >>
> >> 1.3.10:
> >> 1989 180504 [main] ls 436 _pinfo::exit: Calling ExitProcess 0
> >> ^^^^^^
> >>
> >> 1.3.11:
> >> 2447 169876 [main] ls 1480 _pinfo::exit: Calling ExitProcess 0
> >> ^^^^^^
> >>
> >> Are you sure you saw a slowdown in this instance? Because this is
> >> pretty puzzling if so. Given the fact that we cleaned up some of
> >> the startup code in cygwin for 1.3.11, the above figures are what
> >> I'd expect.
> >
> >Pretty sure. I also ran 'time ls' right after I ran 'strace ls'. Here are
> those
> >results, which I meant to post earlier but forgot. I took these times while
> I
> >was logged in to my domain account and connected to my home network, which
> is
> >when I see the slowdown.
>
> But, the strace contains its own version of the time command and it is saying
> that the straced 1.3.11 version is not slow.
>
> Out of curiousity, does mounting your /bin and /usr/bin directories with the
> -X
> option "fix" anything?
>
> mount -f -b -X c:/cygwin/bin /bin
> mount -f -b -X c:/cygwin/bin /usr/bin
>
> where c:\cygwin is the location of your root, of course.
I'll try this when I get home this evening. That would certainly be different,
though, because they're not mounted that way now.
>
> >So for whatever reason, the real time required for ls to do essentially
> nothing
> >is 2.3 seconds longer with the 1.3.11 dll than it is with 1.3.10, while user
> >and system times are more-or-less unchanged.
>
> Sounds like it is timing out trying to contact something like a domain
> controller,
> maybe?
>
Well, that's what I assumed, but what's trying to contact the controller? It's
not ls -- I just used that as an example. I see similar results with *all*
commands. When I was looking through the strace output, there seems to be a
function called internal_getlogin, which appears to be operating somewhat
differently between 1.3.10 and 1.3.11. Under 1.3.10, internal_getlogin appears
to call GetLoginName() along with some other functions to determine my login
name, domain, login server, etc. I don't see similar output in the 1.3.11
strace log. Could that be significant?
--Rick
__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Bug reporting: http://cygwin.com/bugs.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
- Raw text -