Fwd: trouble running tests for 0.7.10

On Sat, Feb 28, 2009 at 7:18 PM, Marcus Lindblom <macke <at> yar.nu> wrote:
> Łukasz Jernaś wrote:
> There is also a system-wide hgrc in /etc/mercurial on unixy systems (at
> least, I've got one in cygwin.)
>
> Which Mercurial version are you running?
>
buildbotslave <at> r17272:/tmp/buildbot$ hg --version
Mercurial Distributed SCM (version 1.0.1)

> Also, can you run the test-suite locally on that slave and see what happens?
> (i.e. 'trial.py buildbot.test.test_vc.Mercurial')

buildbotslave <at> r17272:/tmp/buildbot$ trial buildbot.test.test_vc.Mercurial
Running 5 tests.
buildbot.test.test_vc
 Mercurial
[--snip--]

-------------------------------------------------------------------------------
Ran 5 tests in 96.671s

PASSED (successes=5)

Maybe it has something todo with usepty=0 on my buildbot?

Regards,
--
Łukasz

Re: Fwd: trouble running tests for 0.7.10

Łukasz Jernaś wrote:
> On Sat, Feb 28, 2009 at 7:18 PM, Marcus Lindblom <macke <at> yar.nu> wrote:
>> Łukasz Jernaś wrote:
>> There is also a system-wide hgrc in /etc/mercurial on unixy systems (at
>> least, I've got one in cygwin.)

Did you check the system-wide hgrc?

>> Which Mercurial version are you running?
>>
> buildbotslave <at> r17272:/tmp/buildbot$ hg --version
> Mercurial Distributed SCM (version 1.0.1)

Fair enough.

>> Also, can you run the test-suite locally on that slave and see what happens?
>> (i.e. 'trial.py buildbot.test.test_vc.Mercurial')
> 
> buildbotslave <at> r17272:/tmp/buildbot$ trial buildbot.test.test_vc.Mercurial
> Running 5 tests.
> buildbot.test.test_vc
>  Mercurial
> [--snip--]
> 
> -------------------------------------------------------------------------------
> Ran 5 tests in 96.671s
> 
> PASSED (successes=5)

Hm. 

Re: Fwd: trouble running tests for 0.7.10

On Sun, Mar 1, 2009 at 8:47 AM, Łukasz Jernaś <deejay1@...> wrote:
> On Sat, Feb 28, 2009 at 7:18 PM, Marcus Lindblom <macke@...> wrote:
>> Also, can you run the test-suite locally on that slave and see what happens?
>> (i.e. 'trial.py buildbot.test.test_vc.Mercurial')
>
> buildbotslave <at> r17272:/tmp/buildbot$ trial buildbot.test.test_vc.Mercurial
> Running 5 tests.
> buildbot.test.test_vc
>  Mercurial
> [--snip--]
[...]
> Maybe it has something todo with usepty=0 on my buildbot?

You can approximate "usepty=0" on the command line by redirecting
standard input, output and error to/from non-tty file descriptors.

Something like the following should work:

trial buildbot.test.test_vc.Mercurial < /dev/null 2>&1 | tee log.txt

If it has something to do with process groups, I think you can add a
"nohup" to the beginning of the command to test that theory. However,
you might need to wrap the whole thing in a "nohup sh -c '...'".

--

-- 
- Charles Lepple

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise

test regression upgrading from twisted-8.1.0 to twisted-8.2.0 (with 0.7.10+)

I got the patch from the Git patch-0.7.10 branch integrated into the
Fink package, but I seem to have run across a regression that cropped
up when I upgraded to Twisted 8.2.0.

Part of the diff between the build logs with twisted-8.1.0 and -8.2.0
(using Python 2.5 both times):

   UploadDirectory
-    testManyDirsManyFiles ...                                              [OK]
+    testManyDirsManyFiles ...                                           [ERROR]
...
+===============================================================================
+[ERROR]: buildbot.test.test_transfer.UploadDirectory.testManyDirsManyFiles
+
+Traceback (most recent call last):
+Failure: twisted.internet.error.ProcessTerminated: A process has
ended with a probable error condition: process ended with exit code 1.

This seems to be repeatable - I have run into this every time that I
build against 8.2.0, which is about 3-4 times so far. I have also
tried it outside the Fink framework (in the Git clone) just to see if
there was another tarball-related issue, but that does not seem to
make a difference.

Where should I be looking for additional error information?

The rest of the build logs look pretty mundane, but I can post them if
necessary.

(As an aside, I am seeing a lot of DeprecationWarnings that did not

Re: test regression upgrading from twisted-8.1.0 to twisted-8.2.0 (with 0.7.10+)

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H

_______________________________________________
Buildbot-devel mailing list
Buildbot-devel@...
https://lists.sourceforge.net/lists/listinfo/buildbot-devel

Re: test regression upgrading from twisted-8.1.0 to twisted-8.2.0 (with 0.7.10+)

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H

_______________________________________________
Buildbot-devel mailing list
Buildbot-devel@...
https://lists.sourceforge.net/lists/listinfo/buildbot-devel

OS X, SVN and Buildbot

Hi all,

I thought I would recount a recent problem I encountered involving OS
X, SVN and Buildbot, to pass on information to others (and elicit
sympathy).

We recently moved our subversion server, and switch authentication
from ssh to https. the Mac OS X buildslave didn't like this. I had the
buildslave running as a LaunchDaemon using the instructions here
http://buildbot.net/trac/wiki/UsingLaunchd. For reasons that aren't
clear, running it as a Daemon meant that it couldn't access the cached
authentication info in ~/username/.subversion, and was requesting a
password.

After banging my head and the internetz against this problem for a few
hours, I came across this info
(http://developer.apple.com/technotes/tn2005/tn2083.html#SECACCESSINGUSERSTATE)
stating that Mac Daemons can't reliably access user's home
directories.

To workaround the problem I set the buildslave up to run as a
LaunchAgent, with that user logging in at startup by default.

As another slight annoyance, Mac OS X 10.5 doesn't like "svn
--non-interactive" when using https (worked great under ssh). I found
several complaints around the internet abou this bug, but ended up
removing it from the Mac buildslave slave/commands.py.

Let me know if this sounds familiar to anyone.

Mark MacVicar

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H

Re: OS X, SVN and Buildbot

This sounds similar to a problem we've encountered. Our Mac buildbot  
slaves become unable to checkout code via CVS and other weird problems  
if they are started via ssh.

If started via Terminal via VNC and that Terminal is left open,  
everything is fine. It's also fine using launchd if the user is set to  
automatically log in.

Other than one of those two things we haven't found a solution to this.

On 2-Mar-09, at 5:32 PM, Mark MacVicar wrote:

> Hi all,
>
> I thought I would recount a recent problem I encountered involving OS
> X, SVN and Buildbot, to pass on information to others (and elicit
> sympathy).
>
> We recently moved our subversion server, and switch authentication
> from ssh to https. the Mac OS X buildslave didn't like this. I had the
> buildslave running as a LaunchDaemon using the instructions here
> http://buildbot.net/trac/wiki/UsingLaunchd. For reasons that aren't
> clear, running it as a Daemon meant that it couldn't access the cached
> authentication info in ~/username/.subversion, and was requesting a
> password.
>
> After banging my head and the internetz against this problem for a few
> hours, I came across this info
> (http://developer.apple.com/technotes/tn2005/tn2083.html#SECACCESSINGUSERSTATE 
> )
> stating that Mac Daemons can't reliably access user's home
> directories.
>
> To workaround the problem I set the buildslave up to run as a
> LaunchAgent, with that user logging in at startup by default.
>
>
> As another slight annoyance, Mac OS X 10.5 doesn't like "svn
> --non-interactive" when using https (worked great under ssh). I found
> several complaints around the internet abou this bug, but ended up
> removing it from the Mac buildslave slave/commands.py.
>
> Let me know if this sounds familiar to anyone.
>
> Mark MacVicar
>
> ------------------------------------------------------------------------------
> Open Source Business Conference (OSBC), March 24-25, 2009, San  
> Francisco, CA
> -OSBC tackles the biggest issue in open source: Open Sourcing the  
> Enterprise
> -Strategies to boost innovation and cut costs with open source  
> participation
> -Receive a $600 discount off the registration fee with the source  
> code: SFAD
> http://p.sf.net/sfu/XcvMzF8H
> _______________________________________________
> Buildbot-devel mailing list
> Buildbot-devel@...
> https://lists.sourceforge.net/lists/listinfo/buildbot-devel

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H

Bazaar tests failure

Hi,
I patched the 0.7.10 rpm package I'm working on to add the bits
missing from the release. so far so good.

Now I have the bazaar tests failing. The first one with:

[ERROR]: buildbot.test.test_vc.Bazaar.testCheckout
Traceback (most recent call last):
  File "/usr/lib64/python2.5/site-packages/twisted/internet/defer.py",
line 576, in _deferGenerator
    result = g.next()
  File "/builddir/build/BUILD/buildbot-0.7.10/buildbot/test/test_vc.py",
line 1796, in _createRepository_1
    yield w; out = w.getResult()
  File "/usr/lib64/python2.5/site-packages/twisted/internet/defer.py",
line 555, in getResult
    self.result.raiseException()
  File "/usr/lib64/python2.5/site-packages/twisted/python/failure.py",
line 259, in raiseException
    raise self.type, self.value, self.tb
exceptions.TypeError: exceptions must be classes, instances, or
strings (deprecated), not type

then all other ones with something like:

[ERROR]: buildbot.test.test_vc.Bazaar.testCheckoutBranch
Traceback (most recent call last):
  File "/usr/lib64/python2.5/site-packages/twisted/internet/defer.py",
line 576, in _deferGenerator
    result = g.next()
  File "/builddir/build/BUILD/buildbot-0.7.10/buildbot/test/test_vc.py",
line 1750, in _createRepository_1
    self.populate(tmp)
  File "/builddir/build/BUILD/buildbot-0.7.10/buildbot/test/test_vc.py",
line 294, in populate
    os.makedirs(os.path.join(basedir, "subdir"))
  File "/usr/lib64/python2.5/os.py", line 171, in makedirs
    mkdir(name, mode)
exceptions.OSError: [Errno 17] File exists:
'/builddir/build/BUILD/buildbot-0.7.10/_trial_temp/test_vc/repositories/archtmp/subdir'
===============================================================================

I'll appreciate some help to fix it (otherwise I'll need to disable
bazaar tests).

Of course, feel free to ask me any additional detail

TIA

Gianluca

--

-- 
Gianluca Sforna

http://morefedora.blogspot.com
http://www.linkedin.com/in/gianlucasforna

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H

Re: OS X, SVN and Buildbot

On Mar 2, 2009, at 2:45 PM, Ben Hearsum wrote:

> This sounds similar to a problem we've encountered. Our Mac buildbot
> slaves become unable to checkout code via CVS and other weird problems
> if they are started via ssh.
>
> If started via Terminal via VNC and that Terminal is left open,
> everything is fine. It's also fine using launchd if the user is set to
> automatically log in.
>
> Other than one of those two things we haven't found a solution to  
> this.

These are all because of the way that OS X handles security sessions.   
If you log in via ssh, start some child processes, then log out, the  
user's session is torn down, disabling certain abilities (like  
DirectoryService authentication) for child processes that linger about.

The best way to use buildbot under OS X is to either a) launch it from  
a terminal window on the slave, or b) create a launchd plist to run  
the slave as.  Over ssh, you can sometimes start the buildbot slave  
with "nohup buildbot start <pathToMaster>" (man nohup for more) but  
even that's not guaranteed to work.

	~ john

=================
support the arts!
www.smadness.com

updated October 12, 2005

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H