Re: cvs2svn terminates invisibly

On 05/30/2014 05:37 PM, Ivan wrote:
> I have a cvs repository of about 18Gb and is about 70 projects that is
> about 13 years old.
>  
> I can convert just about any one project with no problem. It goes thru
> the various passes and eventually finishes.
>  
> The problem is some predictable but unpredictable termination.
>  
> I try to convert the whole repository (70 projects), I see the script
> running in the process table; but then the conversion (pass 1) hits the
> very last single file in the repository and stops (with no error message
> or status message).
> [...]
> Next, I try to convert a subset of the repository (20 projects). The
> script then terminates at the exact samespot, and on the same file.  I
> tried to remove that last file and try the conversion, but the cvs2svn
> script terminates on the file before that which was deleted.
> [...]
> my basic command is:
> /usr/bin/cvs2svn  -vv  --tmpdir=/logs/ep/cvs_arch/mpArch
> --dumpfile=/logs/myspot/cvs_arch/mpArch/myProj_cvs2svn.dump  
> --encoding='utf8'  --encoding='latin1'  --encoding='ascii'  ./repo1

A multiproject conversion is qualitatively different than a single
project conversion.  Branches and tags are handled differently, and the
output Subversion repository has a different layout.

To do a multiproject conversion, you have to use the options-file method
[1,2] to start the conversion; otherwise you won't get the correct
layout in the resulting Subversion repository.

I don't know exactly what problem you are running into, but probably the
mismatch between your CVS repository layout and the Subversion project
layout is causing cvs2svn to do an enormous amount of unnecessary work,
causing it either to take forever or overflow memory or something.

> I am using rhel 5, with 4-cpu and 16GB ram and about 30Gb free disk
> space.  No messages ever show up in the cvs2svn log or in
> /var/log/messages, about any diskspace issue.

For such a large conversion, make sure you are using a computer that is
running in 64-bit mode so that the process doesn't overflow its address
space.

> I do wonder:  If I convert the repository in pieces, like 7 dumpfiles of
> 10-projects each, will my import of 7 dumpfiles into svn still retain
> the CVS data intact  ??

Yes, but the commits will not be in chronological order.  This approach
is discussed in the FAQ [3].

Michael

[1] http://cvs2svn.tigris.org/cvs2svn.html#cmd-vs-options
[2] http://cvs2svn.tigris.org/faq.html#onetoone
[3] http://cvs2svn.tigris.org/faq.html#oneatatime

-- 
Michael Haggerty
[email protected]
http://softwareswirl.blogspot.com/

------------------------------------------------------
http://cvs2svn.tigris.org/ds/viewMessage.do?dsForumId=1670&dsMessageId=3079621

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