Hello, B.!
Friday, November 14, 2003, 12:33:16 AM, you wrote:
Ok, I'll not discuss cvs2sbn.pl after this message here, you are
right, I need different mailing list for it, and I'll do it.
>> I have two strong opinions regarding cvs2svn.pl:
>> 1. It is carelessly named, as it will inevitably be confused with
>> cvs2svn.py. My recommendation is that the author be a little
>> more creative, and a lot more sensitive to the fact that
>> scripts are often referred to without their
>> interpreter-indicating filename extensions. In my mind,
>> "cvs2svn" *means* cvs2svn.py -- don't mess with my head.
I named my *project* (not only script, but whole project)
cvs2svn.pl, distribution archive named `cvs2svn.pl' (not simple
`cvs2svn'), you could not see it, but directory on my SVN
repository named `cvs2svn.pl'. Full name of my project is cvs2svn.pl.
>>
>> 2. It is not part of Subversion itself, does not live in
>> Subversion's tree, and is not distributed with Subversion. So
>> please, while it's fine to announce new releases here, kindly
>> take further discussion to a list dedicated to this entirely
>> different product.
Good point. I agree. I'll prepare list form my script today.
BWF> I've got another one right here:
BWF> 3. cvs2svn.py is the result of hundreds of man-hours and a helluva lot
BWF> of careful testing of a helluva lot of tricky edge cases, gargantuan
BWF> repositories, and busted RCS files. Why should I trust my
BWF> *historical*data* to a script that has no visible test suite?
Visible test suit is FreeBSD repository at point of version 4.7. Is
it enough? Some parts of it could not be converted with my script
due to FATAL errors (like revisions 1.1.2.1, which is format
violations, etc). But parts, which could be PARSED converted well
with some help from user, and FAILED without this help.
Also, I've tested my script on test suite from cvs2svn.py. Only three
errors was found. And they are fixed now (and fixes will be in next
release).
Also, my script is over-paranoid by default *especially* to avoid
historical data loss. It is like a cat: ``when you are not sure,
wash yourself''. My script, when it is not sure, asks user. It will
complain about DOZENS situations, when it will works properly with
simple ignoring occurred ``error''. I think, that better here will
be second try with options from user, than some ambiguous decision
will be done, because `any of these elements could be selected, this
element is first in array, lets select it'.
And cvs2svn.py with "big visible test suite" munch many parts of
FreeBSD repository (and repositories of my project!) without any
error, and after that dump are broken or created from this dump
repository contains bad branches, etc.
Also, I don't see how I could help cvs2svn.py -- I could not give
any hints to it, but --only-trunk, which is not a solution in most
cases :(
Lev Serebryakov
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Fri Nov 14 08:43:08 2003