Mark Phippard | 1 Sep 01:45 2007
Picon

Re: Let's set a date/goal for the 1.5 Branch

On 8/31/07, Hyrum K. Wright <hyrum_wright <at> mail.utexas.edu> wrote:

> Daniel Rall wrote:
> > On Fri, 31 Aug 2007, Mark Phippard wrote:
> > ...
> >> 3.  I am assuming is fairly stable already, known issues not with
> >> standing.  I would hope we would branch and cut RC1 all in a short
> >> time frame.  At least that is what I proposing.
> >
> > This is what I groked from Mark's original message -- get ready to cut
> > an RC by mid-October (which implies a branch).  If we have a lot messy
> > outstanding issues remaining at this time, there's no point in cutting
> > an RC, and thus no point in branching (due to porting overhead,
> > testing benefits, etc.).
> >
> > So, as Greg said, the issue list defines the RC point.  And as Mark
> > suggested, let's shoot to knock all the issues off that list by
> > mid-October.
>
> As part of this process, it would be good to cull the 1.5-consider herd,
> and either move stuff postpone it to 1.6 (preferable) or move it to 1.5.
>  Not that I'm volunteering, just something more that should be done.

Well if we did not already put them in the list, odds are this is not
the best time, but we can always see where things stand in a couple of
weeks in terms of progress.  I suspect there are several items
currently in the 1.5 milestone we will want to postpone, rather than
adding new stuff.

Certainly at some point we need to change those milestones though, so yes.
(Continue reading)

David James | 1 Sep 02:10 2007
Picon

Re: Merging Ctypes Python Branch

On 8/30/07, Eric Gillespie <epg <at> pretzelnet.org> wrote:
> "Sage La Torra" <sagelt <at> gmail.com> writes:
>
> > As of last week, David James and I think that the Ctypes Python
> > Binding Branch is ready for merging, or at least a thorough review in
> > preparation for merging. The branch is available at
> > http://svn.collab.net/repos/svn/branches/ctypes-python-bindings/.
>
> As a first step to reviewing, I tried to build it.  Currently,
> you assume everything is installed on the linker search path, but
> this is frequently not the case.  So, I changed autogen.py and
> ctypesgen/wrap.py to take -L, -R, and --rpath options (as well as
> the -Wl,-prefixed forms) to specify libdirs to search.
>
> I also changed autogen.py to use apr-1-config --ldflags --link-ld
> so that ensuring autogen.py has the right apr-1-config is all a
> user needs to do if all svn stuff is in the same libdir.

Hi Eric,

Currently, autogen.py assumes that both apr and apr-util are installed
under the same prefix, according to the output of apr-1-config
--prefix. Further, you can specify where Subversion is installed by
using the "--prefix" option for autogen.py.

We add these directories to the default linker path in autogen.py by
setting the LIBRARY_PATH environment variable.

On my system, Subversion is installed in $HOME. APR is installed in
/usr/local/apache2. To configure the ctypes Python bindings, I run the
(Continue reading)

masaru tsuchiyama | 1 Sep 02:39 2007
Picon

Re: Merging Ctypes Python Branch

Hello

I tested it on Windows XP and got the following error.

D:\svnwork\vc6\ctypes-python-bindings>autogen.py
None
Traceback (most recent call last):
  File "D:\svnwork\vc6\ctypes-python-bindings\autogen.py", line 114, in <module>
    library_path) = get_apr_config()
  File "D:\svnwork\vc6\ctypes-python-bindings\autogen.py", line 54, in
get_apr_config
    raise Exception("Cannot find apr-1-config or apr-config. Please specify\n"
Exception: Cannot find apr-1-config or apr-config. Please specify
the full path to your apr-1-config or apr-config script
using the --apr-config option.

I think autogen.py should have the following options like as gen-make.py.
 --with-apr=DIR
 --with-apr-util=DIR
 --with-httpd=DIR

Regards.
Masaru.

2007/8/31, Sage La Torra <sagelt <at> gmail.com>:
> As of last week, David James and I think that the Ctypes Python
> Binding Branch is ready for merging, or at least a thorough review in
> preparation for merging. The branch is available at
> http://svn.collab.net/repos/svn/branches/ctypes-python-bindings/.
>
(Continue reading)

Justin Erenkrantz | 1 Sep 02:39 2007

Re: Merging Ctypes Python Branch

On 8/31/07, David James <james <at> cs.toronto.edu> wrote:
> Currently, autogen.py assumes that both apr and apr-util are installed
> under the same prefix, according to the output of apr-1-config
> --prefix. Further, you can specify where Subversion is installed by
> using the "--prefix" option for autogen.py.

You should be using apu-1-config to ask apr-util where it lives - not
assuming that they live in the same location.  And, the 'standard'
flags should be --with-apr and --with-apr-util.  -- justin
Justin Erenkrantz | 1 Sep 02:41 2007

Re: Merging Ctypes Python Branch

On 8/31/07, masaru tsuchiyama <m.tmatma <at> gmail.com> wrote:
> I tested it on Windows XP and got the following error.
>
> D:\svnwork\vc6\ctypes-python-bindings>autogen.py
> None
> Traceback (most recent call last):
>   File "D:\svnwork\vc6\ctypes-python-bindings\autogen.py", line 114, in <module>
>     library_path) = get_apr_config()
>   File "D:\svnwork\vc6\ctypes-python-bindings\autogen.py", line 54, in
> get_apr_config
>     raise Exception("Cannot find apr-1-config or apr-config. Please specify\n"
> Exception: Cannot find apr-1-config or apr-config. Please specify
> the full path to your apr-1-config or apr-config script
> using the --apr-config option.

FWIW, APR does not provide any Win32 config scripts.  -- justin
masaru tsuchiyama | 1 Sep 02:51 2007
Picon

Re: Merging Ctypes Python Branch

Hello

> I think autogen.py should have the following options like as gen-make.py.
>  --with-apr=DIR
>  --with-apr-util=DIR
>  --with-httpd=DIR
But If autogen.py will be merged to gen-make.gy when merging the
branch to trunk,
it may not be a problem.

Regards.
Masaru.
Eric Gillespie | 1 Sep 03:14 2007

Re: Merging Ctypes Python Branch

"David James" <james <at> cs.toronto.edu> writes:

> We add these directories to the default linker path in autogen.py by
> setting the LIBRARY_PATH environment variable.

I don't think that's not enough.  The libdirs must be on the
system search path (i.e. listed in /etc/ld.so.conf on Linux) or
listed in the LD_LIBRARY_PATH environment variable.

> Could you try the patched version of ctypesgen on your system and see
> if it helps? You'll need to first update ctypesgen by typing ( cd
> ctypesgen && svn up ). I'm particularly interested to see if this
> works on OS X, which has a different linker and might require us to
> play with different options.

But I'm not entirely sure I understand these things correctly, so
I reverted and updated in a new working copy and tried again,
this time using the --prefix option.  Same error:

OSError: apr-1: cannot open shared object file: No such file or directory

But, you keep referring to your patch, which I do not see.  Is
your patch committed, so updating is sufficient?

> P.S. If my patch solves your problem, are there still parts of your
> patch that you'd still like me to commit? Let me know.

Yes, apr-*-config scripts provide enough information for autogen
to work without any options at all, and my patch was mainly about
getting that information, while also allowing users to specify
(Continue reading)

Mark Phippard | 1 Sep 03:19 2007
Picon

blame_tests 11 failing

This has been happening for a while.  Buildbots do not seem to have it
so maybe can see something wrong with my build?

fsfs or bsb via  ra_local gets same error:

./blame_tests.py 11 --verbose
CMD: svnadmin create svn-test-work/local_tmp/repos --bdb-txn-nosync
<TIME = 0.129432>
CMD: svn import --username jrandom --password rayjandom -m "Log
message for revision 1." svn-test-work/local_tmp/greekfiles
file:///Users/mphippard/work/svn-trunk/subversion/tests/cmdline/svn-test-work/local_tmp/repos
--config-dir /Users/mphippard/work/svn-trunk/subversion/tests/cmdline/svn-test-work/local_tmp/config
<TIME = 0.165843>
CMD: svnadmin create svn-test-work/repositories/blame_tests-11
--bdb-txn-nosync <TIME = 0.105889>
CMD: svnadmin load --force-uuid --quiet
svn-test-work/repositories/blame_tests-11 <TIME = 0.453543>
CMD: svn co file:///Users/mphippard/work/svn-trunk/subversion/tests/cmdline/svn-test-work/repositories/blame_tests-11
svn-test-work/working_copies/blame_tests-11 --config-dir
/Users/mphippard/work/svn-trunk/subversion/tests/cmdline/svn-test-work/local_tmp/config
<TIME = 0.501636>
subversion/libsvn_subr/path.c:344: failed assertion
`is_canonical(path->data, path->len)'
CMD: /Users/mphippard/work/svn-trunk/subversion/svn/svn terminated by signal 6
CMD: svn blame -g
svn-test-work/working_copies/blame_tests-11/trunk/A/upsilon
--config-dir /Users/mphippard/work/svn-trunk/subversion/tests/cmdline/svn-test-work/local_tmp/configEXCEPTION:
SVNProcessTerminatedBySignal
FAIL:  blame_tests.py 11: don't look for merged files out of range

(Continue reading)

Vincent Lefevre | 1 Sep 03:32 2007

Silent working copy corruption after "svn revert"

After a "svn revert", some file was reverted to an old version:

$ svn --version
svn, version 1.4.4 (r25188)
   compiled Aug 27 2007, 23:00:48
[...]
$ pwd
/Users/vinc17/wd/macosx/dports/gnome/liferea
$ svn ls -v
  18910 lefevre          1752 Sep 01 01:43 Portfile
$ svn st -u
Status against revision:  18910
$ svn info Portfile
Path: Portfile
Name: Portfile
URL: svn+ssh://ay/home/lefevre/svn/macosx/dports/gnome/liferea/Portfile
Repository Root: svn+ssh://ay/home/lefevre/svn
Repository UUID: 99759db8-4ec0-0310-8bf9-df86780d22d8
Revision: 18910
Node Kind: file
Schedule: normal
Last Changed Author: lefevre
Last Changed Rev: 18910
Last Changed Date: 2007-09-01 01:43:40 +0200 (Sat, 01 Sep 2007)
Text Last Updated: 2007-09-01 02:21:05 +0200 (Sat, 01 Sep 2007)
Checksum: 7c275625bda2290ce15a71447bc9f52c
$ grep 'Id\|version' Portfile
# $Id: Portfile 18910 2007-08-31 23:43:40Z lefevre $
version             1.0.27
$ svn cat -r18910 Portfile | grep 'Id\|version'
(Continue reading)

Eric Gillespie | 1 Sep 04:01 2007

Re: Merging Ctypes Python Branch

Eric Gillespie <epg <at> pretzelnet.org> writes:

> But, you keep referring to your patch, which I do not see.  Is
> your patch committed, so updating is sufficient?

I have time to look at this again now.  If I add the correct -L
flag to your gcc command, it works.  So, I think we need to merge
my patch with your change: take the libdir option stuff from
mine, drop my load_library change, and then your new find_library
should make things work.

--

-- 
Eric Gillespie <*> epg <at> pretzelnet.org

Gmane