Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Re: Parrot 0.4.0 "Luthor" Released!

4 views
Skip to first unread message

Michael Cummings

unread,
Dec 6, 2005, 6:38:52 PM12/6/05
to Leopold Toetsch, perl6-i...@perl.org
On Tue, 2005-12-06 at 21:13 +0100, Leopold Toetsch wrote:
> > If it helps - the t/perl/manifest test hangs on test 3. Looking at the
> > source, looks like someone intended a diag message to warn that it
> > would
> > appear to hang, but the diag doesn't get printed out until after the
> > fact.
>
> The test takes too much time, but really should finish. Depends of
> course on your hardware, how long it takes.

(sorry if this is a double post for you, I had a bad cc line for
perl6-internals)

Problem is the diag message warning users that it will take a while (and
it does - on a p4 with 2 gigs of ram i gave up after a few minutes, but
didn't get a chance to see where/why it was so slow) is that it doesn't
print before or during the 3rd test - so they have no idea why this
particular test is stuck or that it should take a while. I should
probably follow this up with some patches or something...(see what I can
do)

~mcummings

Leopold Toetsch

unread,
Dec 6, 2005, 8:18:02 PM12/6/05
to Michael Cummings, perl6-i...@perl.org

On Dec 7, 2005, at 0:29, Michael Cummings wrote:

> Problem is the diag message warning users that it will take a while
> (and
> it does - on a p4 with 2 gigs of ram i gave up after a few minutes

On that hardware it certainly shouldn't need minutes to finish. It's
just strange:

$ time perl -Ilib t/perl/manifest.t
1..3
ok 1 - MANIFEST exists
ok 2 - MANIFEST.SKIP exists
# this may take a while...
ok 3 - manicheck()

real 0m6.796s

This is on a PowerBook G4 with 1.2 GHz - certainly not a fast system,
~7 seconds for the test is a pita (if you run it really a lot per day,
but else not a problem)

With ...

$ time perl t/harness t/perl/manifest.t
t/perl/manifest....ok 2/3# this may take a while...

... I see the diag message too, but it may be OS dependent ... hmm ...
no problem on x86 linux (and a lot faster (1 sec) than on OS/X).

Which OS are you using?

> ~mcummings

leo

Michael Cummings

unread,
Dec 6, 2005, 8:26:25 PM12/6/05
to perl6-i...@perl.org
I realize I'm talking to myself at this point (last post, promise), but
my last message failed to explain the paste at the top. On a box without
svk, using the 0.4.0 released tarball, all is fine (t/perl/manifest.t is
skipped/failed quickly). However, if a user has svk installed and
attempts to run make test, this test hangs indefinitely, presumably svk
getting confused. Maybe a (-d .svn) should be checked first to be sure
we aren't using the release tarball instead of a subversion checkout?

Michael Cummings

unread,
Dec 6, 2005, 8:13:13 PM12/6/05
to Leopold Toetsch, perl6-i...@perl.org
On Tue, 2005-12-06 at 18:38 -0500, Michael Cummings wrote:

t/perl/manifest......................Can't exec "svk": No such file or
directory at t/perl/manifest.t line 38.
ok
1/3 skipped: Not a working copy

Not a patch, but a better explanation of the problem. There is a false
assumption in the t/perl/manifest.t test that svk is installed only if
your dealing with parrot from svn :) svk ls . on a directory that isn't
actually checked out from svn (say, on a dir from the tarball that was
released as 0.4.0 publicly) results in a long, long, long pause...


Jerry Gay

unread,
Dec 6, 2005, 8:19:00 PM12/6/05
to Michael Cummings, Leopold Toetsch, perl6-i...@perl.org
good, that helps. i'm not an svk user, so i can't really test a patch,
but there are svk users here--who, if so inclined, can use this
description to create a proper fix. of course, don't let that stop you
from creating a patch yourself ;)

thanks for reporting.
~jerry

Michael Cummings

unread,
Dec 6, 2005, 8:38:15 PM12/6/05
to Leopold Toetsch, perl6-i...@perl.org
On Wed, 2005-12-07 at 02:18 +0100, Leopold Toetsch wrote:
> On Dec 7, 2005, at 0:29, Michael Cummings wrote:
>
> On that hardware it certainly shouldn't need minutes to finish. It's
> just strange:

> This is on a PowerBook G4 with 1.2 GHz - certainly not a fast system,

> ~7 seconds for the test is a pita (if you run it really a lot per day,
> but else not a problem)
>
> With ...
>
> $ time perl t/harness t/perl/manifest.t
> t/perl/manifest....ok 2/3# this may take a while...
>
> ... I see the diag message too, but it may be OS dependent ... hmm ...
> no problem on x86 linux (and a lot faster (1 sec) than on OS/X).
>
> Which OS are you using?

Gentoo linux (ask away - I maintain the installation of parrot for
gentoo - stock make/make install sequence, no add-on patches, etc.)

Follow up message narrows this down I think - has to do with not using a
subversion checkout of parrot, but having svk on the box.


Leopold Toetsch

unread,
Dec 7, 2005, 7:05:36 AM12/7/05
to Michael Cummings, perl6-i...@perl.org

Thanks for the detailed analyzes. I've now cleaned up the manifest.t
(r10386). It turned out that the whole check for svn/svk is absolutely
not needed in the test.

leo

Chromatic

unread,
Dec 7, 2005, 12:05:59 PM12/7/05
to Michael Cummings, perl6-i...@perl.org

Would that have helped? svk doesn't use .svn directories.

-- c

0 new messages