Got a recipe for you - cvs2svn problem

Just to add another data point...

If you restrict the conversion deeper in the cvs tree, you
can see
where it breaks.  If I only convert the last level
directory, it
works:

[ebrownnfs01 working]# cvs2svn -s repo01 \
  jboss-common/src/main/org/jboss/util/stream
[ebrownnfs01 working]# svn ls \
  file://repo01/branches/Branch_4_0 | grep IMarsh
---- no file, as it should be ----
[ebrownnfs01 working]# svn log
file://repo01/branches/Branch_4_0 | head -5
------------------------------------------------------------
------------
r239 | csuconic | 2006-03-28 23:16:05 -0500 (Tue, 28 Mar
2006) | 2 lines

JBAS-3025 - Removing dependency on commons/IMarshalledValue
[ebrownnfs01 working]#

See htt
p://download.jboss.org/misc/cvs2svn.log.01.txt for
details,
including cvs2svn -v output.

Now, if I increase the depth up one level, the file show up
in the
Branch_4_0 where it should not (and is not in CVS):

[ebrownnfs01 working]# cvs2svn -s repo02 \
  jboss-common/src/main/org/jboss/util/stream
[ebrownnfs01 working]# svn ls \
file://repo02/branches/Branch_4_0/stream/ | grep IMars
IMarshalledValue.java
---- Doh! we see the file, as it should NOT be ----
[ebrownnfs01 working]# svn log
file://repo02/branches/Branch_4_0/stream/ | head -5
------------------------------------------------------------
------------
r951 | (no author) | 2006-05-10 22:23:22 -0400 (Wed, 10 May
2006) | 1 line

This commit was manufactured by cvs2svn to create branch
'Branch_4_0'.
------------------------------------------------------------
------------
r906 | csuconic | 2006-03-28 23:16:05 -0500 (Tue, 28 Mar
2006) | 2 lines
[ebrownnfs01 working]#

See htt
p://download.jboss.org/misc/cvs2svn.log.02.txt for
details,
including cvs2svn -v output.


> From: Damon Sicore <damon.sicorejboss.com>
> Date: Thu, 1 Jun 2006 19:26:55 -0500
> To: <userscvs2svn.tigris.org>
> Subject: Got a recipe for you - cvs2svn problem
> 
> cvs2svn g33k5:
> 
> I've a recipe for a cvs2svn problem you can download
here:
> 
> http://da
mon.sicore.org/jboss/recipe.tar.gz
> 
> Basically, what happens is I get a file in a resulting
svn repo branch that 
> shouldn't exist (and doesn't in the cvs branch
equivalent).  The 'doIt.sh'
> script will highlight the problem.
> 
> Any help appreciated.
> 
> Sincerely,
> Damon
> 
> ---
> dsicorejboss.org    JBoss Labs Lead     214-883-6733
> 36FD 368C 38E4 2DA2 6E2C  C0A6 FA7F 8A87 EA10 65A9
> Public Key:  http://keys.sicore.o
rg/jboss.txt

-- 
Eric Brown | Director of IT | JBoss, Inc.
GPG: 7573 8AB7 DD77 5CBC 48FD  7C90 1D68 5310 D06B 7679
(o) +1.404.442.2033 (f) +1.404.442.2034 (c) +1.404.964.6062



------------------------------------------------------------
---------
To unsubscribe, e-mail: users-unsubscribecvs2svn.tigris.org
For additional commands, e-mail: users-helpcvs2svn.tigris.org

Damon Sicore wrote:
> I've a recipe for a cvs2svn problem you can download
here:
> 
> http://da
mon.sicore.org/jboss/recipe.tar.gz
> 
> Basically, what happens is I get a file in a resulting
svn repo branch
> that  shouldn't exist (and doesn't in the cvs branch
equivalent).  The
> 'doIt.sh' script will highlight the problem.
> 
> Any help appreciated.

Thanks very much for the excellent test script.  I just
verified that
the same problem appears in cvs2svn trunk.

For the record, the conversion outputs the following
warnings, including
one referring to the affected file.  (The other two files
appear in
Branch_4_0 in both CVS and svn, so the same bug would
probably not be
expected.):

WARNING: Attempt to set timestamp of revision 1.2.2.2 on
file
jboss-common/src/main/org/jboss/util/collection/CollectionsU
til.java to
time 1135107142, which is before previous the time of
revision 1.2.2.1
(1135107143):
WARNING: Time set to 1135107144
WARNING: Attempt to set timestamp of revision 1.1.2.2 on
file
jboss-common/src/main/org/jboss/util/propertyeditor/StringEd
itor.java to
time 1122496896, which is before previous the time of
revision 1.1.2.1
(1122496897):
WARNING: Time set to 1122496898
WARNING: Attempt to set timestamp of revision 1.1.2.2 on
file
jboss-common/src/main/org/jboss/util/stream/IMarshalledValue
.java to
time 1134677906, which is before previous the time of
revision 1.1.2.1
(1134677907):
WARNING: Time set to 1134677908

This is clearly a cvs2svn bug, though it is likely related
to some clock
skew in your repository.  You *might* be able to fix this
specific
conversion problem by hand-editing the
IMarshalledValue.java,v file to
fix up the data sequence, though of course without further
analysis
there is no way to be sure that similar problems don't
appear elsewhere.

Michael

------------------------------------------------------------
---------
To unsubscribe, e-mail: users-unsubscribecvs2svn.tigris.org
For additional commands, e-mail: users-helpcvs2svn.tigris.org

> This is clearly a cvs2svn bug, though it is likely
related to some clock
> skew in your repository.  You *might* be able to fix
this specific
> conversion problem by hand-editing the
IMarshalledValue.java,v file to
> fix up the data sequence, though of course without
further analysis
> there is no way to be sure that similar problems don't
appear elsewhere.

I tried that, and it does in fact clear up the warning
message, but
the root problem persists.  It seems to be unrelated.
 
> Michael
> 
>
------------------------------------------------------------
---------
> To unsubscribe, e-mail: users-unsubscribecvs2svn.tigris.org
> For additional commands, e-mail: users-helpcvs2svn.tigris.org
> 


------------------------------------------------------------
---------
To unsubscribe, e-mail: users-unsubscribecvs2svn.tigris.org
For additional commands, e-mail: users-helpcvs2svn.tigris.org

So what is it about this file:

h
ttp://damon.sicore.org/jboss/IMarshalledValue.java,v

That triggers IMarshalledValue.java to show up in the
Branch_4_0 in  
SVN ?


On Jun 2, 2006, at 5:37 AM, Michael Haggerty wrote:

> Damon Sicore wrote:
>> I've a recipe for a cvs2svn problem you can
download here:
>>
>> http://da
mon.sicore.org/jboss/recipe.tar.gz
>>
>> Basically, what happens is I get a file in a
resulting svn repo  
>> branch
>> that  shouldn't exist (and doesn't in the cvs
branch equivalent).   
>> The
>> 'doIt.sh' script will highlight the problem.
>>
>> Any help appreciated.
>
> Thanks very much for the excellent test script.  I just
verified that
> the same problem appears in cvs2svn trunk.
>
> For the record, the conversion outputs the following
warnings,  
> including
> one referring to the affected file.  (The other two
files appear in
> Branch_4_0 in both CVS and svn, so the same bug would
probably not be
> expected.):
>
> WARNING: Attempt to set timestamp of revision 1.2.2.2
on file
> jboss-common/src/main/org/jboss/util/collection/ 
> CollectionsUtil.java to
> time 1135107142, which is before previous the time of
revision 1.2.2.1
> (1135107143):
> WARNING: Time set to 1135107144
> WARNING: Attempt to set timestamp of revision 1.1.2.2
on file
> jboss-common/src/main/org/jboss/util/propertyeditor/ 
> StringEditor.java to
> time 1122496896, which is before previous the time of
revision 1.1.2.1
> (1122496897):
> WARNING: Time set to 1122496898
> WARNING: Attempt to set timestamp of revision 1.1.2.2
on file
>
jboss-common/src/main/org/jboss/util/stream/IMarshalledValue
.java to
> time 1134677906, which is before previous the time of
revision 1.1.2.1
> (1134677907):
> WARNING: Time set to 1134677908
>
> This is clearly a cvs2svn bug, though it is likely
related to some  
> clock
> skew in your repository.  You *might* be able to fix
this specific
> conversion problem by hand-editing the
IMarshalledValue.java,v file to
> fix up the data sequence, though of course without
further analysis
> there is no way to be sure that similar problems don't
appear  
> elsewhere.
>
> Michael

------------------------------------------------------------
---------
To unsubscribe, e-mail: users-unsubscribecvs2svn.tigris.org
For additional commands, e-mail: users-helpcvs2svn.tigris.org

We would like to contract out one of the cvs2svn developers
to get  
this bug fixed.   This is a big deal for us.

Anyone interested?

On Jun 2, 2006, at 10:55 AM, Damon Sicore wrote:

> So what is it about this file:
>
> h
ttp://damon.sicore.org/jboss/IMarshalledValue.java,v
>
> That triggers IMarshalledValue.java to show up in the
Branch_4_0 in  
> SVN ?
>
>
> On Jun 2, 2006, at 5:37 AM, Michael Haggerty wrote:
>
>> Damon Sicore wrote:
>>> I've a recipe for a cvs2svn problem you can
download here:
>>>
>>> http://da
mon.sicore.org/jboss/recipe.tar.gz
>>>
>>> Basically, what happens is I get a file in a
resulting svn repo  
>>> branch
>>> that  shouldn't exist (and doesn't in the cvs
branch  
>>> equivalent).  The
>>> 'doIt.sh' script will highlight the problem.
>>>
>>> Any help appreciated.
>>
>> Thanks very much for the excellent test script.  I
just verified that
>> the same problem appears in cvs2svn trunk.
>>
>> For the record, the conversion outputs the
following warnings,  
>> including
>> one referring to the affected file.  (The other two
files appear in
>> Branch_4_0 in both CVS and svn, so the same bug
would probably not be
>> expected.):
>>
>> WARNING: Attempt to set timestamp of revision
1.2.2.2 on file
>> jboss-common/src/main/org/jboss/util/collection/ 
>> CollectionsUtil.java to
>> time 1135107142, which is before previous the time
of revision  
>> 1.2.2.1
>> (1135107143):
>> WARNING: Time set to 1135107144
>> WARNING: Attempt to set timestamp of revision
1.1.2.2 on file
>>
jboss-common/src/main/org/jboss/util/propertyeditor/ 
>> StringEditor.java to
>> time 1122496896, which is before previous the time
of revision  
>> 1.1.2.1
>> (1122496897):
>> WARNING: Time set to 1122496898
>> WARNING: Attempt to set timestamp of revision
1.1.2.2 on file
>>
jboss-common/src/main/org/jboss/util/stream/IMarshalledValue
.java to
>> time 1134677906, which is before previous the time
of revision  
>> 1.1.2.1
>> (1134677907):
>> WARNING: Time set to 1134677908
>>
>> This is clearly a cvs2svn bug, though it is likely
related to some  
>> clock
>> skew in your repository.  You *might* be able to
fix this specific
>> conversion problem by hand-editing the
IMarshalledValue.java,v  
>> file to
>> fix up the data sequence, though of course without
further analysis
>> there is no way to be sure that similar problems
don't appear  
>> elsewhere.
>>
>> Michael
>
>
------------------------------------------------------------
---------
> To unsubscribe, e-mail: dev-unsubscribecvs2svn.tigris.org
> For additional commands, e-mail: dev-helpcvs2svn.tigris.org
>

------------------------------------------------------------
---------
To unsubscribe, e-mail: users-unsubscribecvs2svn.tigris.org
For additional commands, e-mail: users-helpcvs2svn.tigris.org

Damon Sicore wrote:
> I've a recipe for a cvs2svn problem you can download
here:
> 
> http://da
mon.sicore.org/jboss/recipe.tar.gz
> 
> Basically, what happens is I get a file in a resulting
svn repo branch
> that  shouldn't exist (and doesn't in the cvs branch
equivalent).  The
> 'doIt.sh' script will highlight the problem.

I just created issue 102 to cover this bug (including an
even-more-simplified reproduction script and some info I've
found out
about the bug):

    
http://cvs2svn.tigris.org/issues/show_bug.cgi?id=102

Followups please to devcvs2svn.tigris.org.

Michael

------------------------------------------------------------
---------
To unsubscribe, e-mail: users-unsubscribecvs2svn.tigris.org
For additional commands, e-mail: users-helpcvs2svn.tigris.org