Re: cvs2svn terminates and hangs

On 06/02/2014 08:32 PM, Ivan wrote:
> I am using cvs2svn on RHEL 5.
> 
> After running for 13 hours my conversion just hangs and terminates with
> no error. It says that it is flushing a cache.
> 
> I let it sit there for about 8 hours and give up.   Why does it stop
> here ?????
> ________
> 27519.086568: repo1/hnweb/web/ui/members/menu/Attic/navigation.jsp,v
> 27519.093897: repo1/hnweb/web/ui/styles/Attic/style.css,v
> 27519.101269: Processed 20244 files
> 27522.084691: Flushing cache for
> RecordTable('/ep/cvs_arch/mpArch/metadata-index.dat')
> ----------------
> 
> that index datfile is only 458k in size.
> this 'flushing cache'  message does not even confirm if pass #1 even
> completed correctly.
> 
> the basic command I use is:
> $cvs2svn  -vv  --dumpfile=/ep/cvs_arch/mpArch/HNonly_cvs2svn.dump
> --encoding='utf8' --encoding='latin1' --encoding='ascii' ./repo1
> 
> I can convert other smaller cvs projects with no problem.  In this
> instance I am trying to convert one giant
> project that has almost 14 years of cvs-commits.  My server is not
> running out of ram either. I checked
> 
> There are no erros in the /var/log/messages file either.

I can't remember ever seeing a problem like this reported.  It seems
mysterious.

What version of Python are you using?  What version of cvs2svn?  How big
is this one project?  Was the project every used with CVSNT (as opposed
to standard CVS)?

There are three classes of problem causes that spring to mind:

1. The problem is due to the size of your repository.  Yours is not the
biggest repository that I have heard of, but maybe some dimension of
your repository (e.g., number of branches/tags, depth of directory tree,
number of files per directory, number of commits) is unusually large.

2. The problem is due to something strange in your repository.  Maybe
one or more files are corrupt?  Maybe there is something strange with
the filesystem (e.g., link cycles).

3. The problem is caused by something strange in your environment.
Maybe you have a malfunctioning version of Python or of some library?
Maybe some quota is being exceeded?

Of course the final cause might be a bug in cvs2svn, but given that
cvs2svn *usually* works, there must be something out of the ordinary
that is triggering the bug.  If we find out what that something is, we
will be a lot closer to figuring out the ultimate cause.

Have you tried the procedure for reducing your problem to a small
reproducible test case [1]?  If this procedure works, it not only speeds
up further testing dramatically, but it sometimes makes the cause of the
problem transparently clear.

Have you tried interrupting the program when it is stuck?  I think that
should cause Python to print a stack trace, which would be very useful.

Have you tried running strace on the program when it is stuck?  This
might show what it is trying to do.

Typically the RecordTable('/ep/cvs_arch/mpArch/metadata-index.dat') will
be flushed many times during a program run.  Do you see the "Flushing"
line earlier in the output, or is this the first such "Flushing" line
with this filename?

Rest assured that cvs2svn has been used for both bigger and older
repositories than yours, and your problem is likely solvable!

Michael

[1] http://cvs2svn.tigris.org/faq.html#testcase

-- 
Michael Haggerty
[email protected]

------------------------------------------------------
http://cvs2svn.tigris.org/ds/viewMessage.do?dsForumId=1670&dsMessageId=3079770

To unsubscribe from this discussion, e-mail: [[email protected]].