make clog slowdown in devel.
 

I noticed something odd..

(18:25:35:davej@gelk:F-9)$ time make clog
make: `clog' is up to date.
0.22user 0.11system 0:00.33elapsed 98%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+52803minor)pagefaults 0swaps
(18:25:40:davej@gelk:F-9)$ cd ../devel/
(18:25:43:davej@gelk:devel)$ time make clog
make: `clog' is up to date.
0.25user 0.15system 0:02.82elapsed 14%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+76947minor)pagefaults 0swaps

The times are consistent across multiple runs. Same command in devel takes
a lot longer. The spec file differences are irrelevant.
(tested by copying the specfile from f9 to devel)

The clog target is just a simple sed & tee, so I'm confused as to why
it takes longer.

Dave

--
http://www.codemonkey.org.uk

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

make clog slowdown in devel.
 

On Fri, 2008-06-20 at 18:29 -0400, Dave Jones wrote:
> I noticed something odd..
[snip]
> The times are consistent across multiple runs. Same command in devel takes
> a lot longer. The spec file differences are irrelevant.
> (tested by copying the specfile from f9 to devel)
>
> The clog target is just a simple sed & tee, so I'm confused as to why
> it takes longer.

I suspect it's due to the fact that we don't have a branch file and so
do more checking for figuring out what branch you're on (which is used
to determine the value of %dist among other things)

Jeremy

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

make clog slowdown in devel.
 

On Fri, Jun 20, 2008 at 10:07:33PM -0400, Jeremy Katz wrote:
> On Fri, 2008-06-20 at 18:29 -0400, Dave Jones wrote:
> > I noticed something odd..
> [snip]
> > The times are consistent across multiple runs. Same command in devel takes
> > a lot longer. The spec file differences are irrelevant.
> > (tested by copying the specfile from f9 to devel)
> >
> > The clog target is just a simple sed & tee, so I'm confused as to why
> > it takes longer.
>
> I suspect it's due to the fact that we don't have a branch file and so
> do more checking for figuring out what branch you're on (which is used
> to determine the value of %dist among other things)

I tried faking one. Didn't change anything..

$ echo F-10 > branch
$ time make clog
make: `clog' is up to date.
0.24user 0.15system 0:02.82elapsed 14%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+74933minor)pagefaults 0swaps

Dave

--
http://www.codemonkey.org.uk

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

make clog slowdown in devel.
 

2008/6/21 Dave Jones <davej@redhat.com>:
> I tried faking one. Didn't change anything..

See the top of Makefile.common. If the current directory is named
"devel", a cvs rlog command is issued to determine whether BRANCH
should be set to "F-10" or "devel". I don't quite grok the logic yet,
but that's the reason for the delay.

- Thomas

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

make clog slowdown in devel.
 

On Saturday 21 June 2008, Dave Jones wrote:
> On Fri, Jun 20, 2008 at 10:07:33PM -0400, Jeremy Katz wrote:
> > On Fri, 2008-06-20 at 18:29 -0400, Dave Jones wrote:
> > > I noticed something odd..
> >
> > [snip]
> >
> > > The times are consistent across multiple runs. Same command in devel
> > > takes a lot longer. The spec file differences are irrelevant.
> > > (tested by copying the specfile from f9 to devel)
> > >
> > > The clog target is just a simple sed & tee, so I'm confused as to why
> > > it takes longer.
> >
> > I suspect it's due to the fact that we don't have a branch file and so
> > do more checking for figuring out what branch you're on (which is used
> > to determine the value of %dist among other things)
>
> I tried faking one. Didn't change anything..
>
> $ echo F-10 > branch
[...]

Makefile.common doesn't seem to be looking at the "branch" file at all to
figure out the branch, instead it does a "pwd | awk -F '/' '{ print $NF }'".

And as Jeremy noted, if the branch is "devel", a subsequent "cvs rlog" is done
before anything else which I'm pretty sure is the slowdown culprit. I wonder
if it would be feasible to modify Makefile.commmon to do that only when
needed so frequently run targets that don't need the branch/dist info
wouldn't be affected. Or just disable the rlog thingy during times when no
early branching for the next release is being done (I suppose that's what it
is all about).

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list