[svk-devel] Re: Still problems using svk w/svn.apache.org

[Jason Dillon]

The plot thickens... after looking into this further... the ASF infra  
folks have allowed access to mirror the entire geronimo/ project tree  
out of the single asf repo, as it was suggested that the copy from  
geronimo/sandbox/... to geronimo/server/... was causing the problem  
below.

So I set a mirror for http://svn.apache.org/repos/asf/geronimo (and  
also tried https too) and now every time I try the sync gets to:

<snip>
...
Committed revision 3253 from revision 159756.
</snip>

And then hangs... does not eat up cpu, does not make any connections  
to the svn server... just stops doing anything.

Re-running the sync produces the same basic results... hits the  
server a few times, then appears to stop doing anything.  It spits  
out only this:

<snip>
Syncing http://svn.apache.org/repos/asf/geronimo
Retrieving log information from 159757 to 514079
</snip>

And after hours and hours of it doing nothing at all, CTRL-C, spits  
this out:

<snip>
Interrupted.
perl(8932) malloc: ***  Deallocation of a pointer not malloced:  
0x30e1000; This could be a double free(), or free() called with the  
middle of an allocated block; Try setting environment variable  
MallocHelp to see tools to help debug
perl(8932) malloc: ***  Deallocation of a pointer not malloced:  
0x30dd000; This could be a double free(), or free() called with the  
middle of an allocated block; Try setting environment variable  
MallocHelp to see tools to help debug
perl(8932) malloc: ***  Deallocation of a pointer not malloced:  
0x3025a00; This could be a double free(), or free() called with the  
middle of an allocated block; Try setting environment variable  
MallocHelp to see tools to help debug
perl(8932) malloc: ***  Deallocation of a pointer not malloced:  
0x310057b; This could be a double free(), or free() called with the  
middle of an allocated block; Try setting environment variable  
MallocHelp to see tools to help debug
Segmentation fault
Interrupted.
</snip>

Looking into the ktrace output, shows the last svn-ish looking  
revision its trying to do something with is 159776.  This revision is  
specific to geronimo, but does not appear to contain anything  
abnormal, or copies from other trees outside of geronimo.

:-(

They have left the http://svn.apache.org/repos/asf/geronimo open to  
test svk stuff for now (normally recursive operations on entire  
project trees are rejected by httpd rules).

--jason


On Mar 1, 2007, at 11:37 PM, Jason Dillon wrote:

> So, svn.apache.org was finally upgrade to svn 1.4.3 today... :-)   
> And though svk sync did get *much, much* further than it did before  
> (when the server was svn 1.3)... it still pukes with RA layer  
> problems... and now it segfaults too... and sometimes spits out  
> these malloc lines over and over and over (sometimes a few hundred):
>
> <snip>
> $ /opt/local/bin/svk sync -a
> Starting to synchronize //mirrors/geronimo/server
> Syncing https://svn.apache.org/repos/asf/geronimo/server
> Retrieving log information from 472725 to 513631
> RA layer request failed: REPORT request failed on 'https:// 
> svn.apache.org/repos/asf/geronimo/server': REPORT of 'https:// 
> svn.apache.org/repos/asf/geronimo/server': Could not read chunk  
> delimiter: Secure connection truncated (https://svn.apache.org)
> Died at /opt/local/lib/perl5/vendor_perl/5.8.8/SVK/Mirror/Backend/ 
> SVNRaPipe.pm line 212.
> perl(4385) malloc: ***  Deallocation of a pointer not malloced:  
> 0x1c6c000; This could be a double free(), or free() called with the  
> middle of an allocated block; Try setting environment variable  
> MallocHelp to see tools to help debug
> perl(4385) malloc: ***  Deallocation of a pointer not malloced:  
> 0x1c6c000; This could be a double free(), or free() called with the  
> middle of an allocated block; Try setting environment variable  
> MallocHelp to see tools to help debug
> perl(4385) malloc: ***  Deallocation of a pointer not malloced:  
> 0x1b2a200; This could be a double free(), or free() called with the  
> middle of an allocated block; Try setting environment variable  
> MallocHelp to see tools to help debug
> perl(4385) malloc: ***  Deallocation of a pointer not malloced:  
> 0x357b; This could be a double free(), or free() called with the  
> middle of an allocated block; Try setting environment variable  
> MallocHelp to see tools to help debug
> Segmentation fault
> </snip>
>
> One of the ASF admins was nice enough to peep into the server logs  
> and reported this:
>
> <snip>
> File not found: revision 472728, path '/geronimo/server/trunk/ 
> modules/geronimo-persistence-jpa10/src/main/java/org/apache/ 
> geronimo/persistence/TransformerWrapper.java'
> the preamble to that is:
> Provider encountered an error while streaming a REPORT response.   
> [404, #0]
> Problem replaying revision  [404, #160013]
> </snip>
>
> I had been using the Mac OSX builds from here:
>
>     http://homepage.mac.com/hiirem/svkbuilds.html
>
> But someone from #asfinfra mentioned something about a new  
> SVK::Mirror and about SVN::Mirror being horrible broken... and  
> since these builds say something about SVN::Mirror, I tried the  
> macports version too (which is what the above output is from).  But  
> both behave in exactly the same way.
>
> Another odd thing I noticed was in the last snip above, its  
> referencing the TransformerWrapper.java in rev 472728... which is  
> very odd... since this change is actually in the next revision 472729.
>
> http://svn.apache.org/viewvc?view=rev&revision=472728
> http://svn.apache.org/viewvc?view=rev&revision=472729
>
> 472728 actually has nothing to do with the Geronimo project (which  
> is what I am trying to use svk with).
>
> I've tried http and https... both fail in the same basic ways,  
> though http pukes with something more like:
>
> <snip>
> $ svk sync -a
> Starting to synchronize //mirrors/geronimo/server
> Syncing http://svn.apache.org/repos/asf/geronimo/server
> Retrieving log information from 472725 to 513639
> RA layer request failed: REPORT request failed on 'http:// 
> svn.apache.org/repos/asf/geronimo/server': REPORT of 'http:// 
> svn.apache.org/repos/asf/geronimo/server': 200 OK (http:// 
> svn.apache.org)
> Can't use an undefined value as an ARRAY reference at /usr/local/ 
> svn/perl/SVK/Mirror/Backend/SVNRaPipe.pm line 250.
> perl(4491) malloc: ***  Deallocation of a pointer not malloced:  
> 0x1cf9a00; This could be a double free(), or free() called with the  
> middle of an allocated block; Try setting environment variable  
> MallocHelp to see tools to help debug
> perl(4491) malloc: ***  Deallocation of a pointer not malloced:  
> 0x1ca7a00; This could be a double free(), or free() called with the  
> middle of an allocated block; Try setting environment variable  
> MallocHelp to see tools to help debug
> perl(4491) malloc: ***  Deallocation of a pointer not malloced:  
> 0x1c45000; This could be a double free(), or free() called with the  
> middle of an allocated block; Try setting environment variable  
> MallocHelp to see tools to help debug
> perl(4491) malloc: ***  Deallocation of a pointer not malloced:  
> 0x1d22f7b; This could be a double free(), or free() called with the  
> middle of an allocated block; Try setting environment variable  
> MallocHelp to see tools to help debug
> Segmentation fault
> </snip>
>
> I even tried to get win32 version (from http:// 
> kevinhlight.googlepages.com/svkwin32) working... but the installer  
> complains that its corrupt (and the md5 does not match).  And  
> anyone who knows me... knows I hate windows... but I was (and still  
> am) willing to use it if helps me debug this problem (though if it  
> would only work on windows... er... well... i'll go shoot myself  
> now :-P).
>
> Anyways... I completely lost on how to get working with svk for  
> Geronimo :-(  I *desperately* want to get svk working to manage  
> branching and some major project layout re-factoring work.  SVK is  
> the closest thing I'm going to get to Perforce when working on ASF  
> stuff... so I really, really, really want to figure out how to make  
> this work... for me and other Geronimo developers.  This is a  
> hugely powerful tool for managing branching and merging with SVN  
> repos and I really, really... really want to figure this out.
>
> Anyone have any idea what might be going on here?  Is this a SVK  
> bug?  A SVN bug?  A corrupt repo?  Server on crack?  User on crack?
>
> Anything would be helpful...
>
> please
>
> /me gets down on his knees
>
> --jason