[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [patch #4573] Fix for keyword expansion problem/mis-feature during c

From: Rahul Bhargava
Subject: Re: [patch #4573] Fix for keyword expansion problem/mis-feature during commit
Date: Mon, 13 Feb 2006 19:06:53 -0800
User-agent: Thunderbird 1.5 (Windows/20051201)

Hello Arthur :

See response below:

Arthur Barrett wrote:
Rahul and Patrick,

In particular your original patch was not suitable for inclusion in
CVSNT (or probably for cvs 1.12) as it has a hardcoded list of keywords,
Like what ?  The keywords[] array in kwfilecmp.c is same as
used by CVS src base. It could be refactored but then kwdiff standalone
utility needs to link with more code ...anyways let me focus on the major points
you raise below ..

which neither CVS nor CVSNT use.  The solution you provided was also
considered to be overcomplex - a much simpler algorithm could be used
very easily by patching the RCS compare routines themselves... all
keywords for example exist on a single line ($Log$ is a special case but
wouldn't be handled by this version either)... In our opinion you could
simply pair a strcmp with a call to expand_keywords and it'd work very
simply, without introducing new untested algorithms into the stable
It seems like you are misunderstanding the algorithm. Its not expanding the
keyword rather un-expanding them to be able to ignore spurious keyword
differences at commit time. Pair of strcmp obviosuly don't make sense as the client may have put bytes in keywords that will not match with anything on server-side. In other words the patch "canonicalizes" the keywords instead of strcmp'ing them. Will be happy to explain the algorithm in more details if you still have confusion. Except for $log$ it takes care of every conceivable differences in keywords that could cause a spurious
commit to occur.

As for subversion being an alternative for commercial software
development, there was a recent thread on that subject ("Missing CVS
Windows Client") that I advise you read carefully since it referenced
cases of repository corruption in SVN with both the file and database
In another thread that I posted on in the CVSNT forum


it appears CVSNT has more stability issues to work out compared to CVS and Subversion.

No one from CVSNT-dev has been able to point to a stable version that can be used for conducting stress tests. In our labs we continuously run stress tests
on various CM backends supported by our products (http://www.wandisco.com).
Many of these stress tests are rigged to crash servers at random points in time. Barring minor issues, CVS and SVN seem to be fairly solid, CVSNT has had more issues.

Recent threads on CVSNT  have posted about repo corruption:


backends.  The CVSNT project endeavours to deliver open source
I am still confused about the CVSNT open source model. For example in this
post - http://www.cvsnt.org/pipermail/cvsnt/2006-February/023724.html
you seem to indicate the commercially supported customers get a different/stable set of bits than the open source bits. The last build that we picked up was from your
company site and you mention in the above post that it was "rejected"  for
commercial support customers yet it was the version available for download as
open source/free version.
versioning tools that facilitates good CM process.  Hence ACL's, Audit,
Mergepoints (Merge Tracking), True Rename (not copy+delete), Unicode
Sure you have an impressive sounding list of features. But it will
be even better if they worked without the stability issues/freezes.

filenames, Unicode Merge etc. None of those features are yet to make it
into SVN:

If you are interested in contacting the CVSNT project then please
contact the newsgroup:

Please read this FAQ before submitting a support request:


Arthur Barrett

-----Original Message-----
From: info-cvs-bounces+arthur.barrett=march-hare.com@nongnu.org
[mailto:info-cvs-bounces+arthur.barrett=march-hare.com@nongnu.org] On
Behalf Of Rahul Bhargava
Sent: 11 February 2006 11:41
To: Patrick Richardson
Cc: bug-cvs@nongnu.org; Info-cvs@nongnu.org
Subject: Re: [patch #4573] Fix for keyword expansion problem/mis-feature
during commit

Sorry to hear that Patrick. Indeed, Subversion implements the same
behavior as the proposed patch.  Subversion allows keywords properties
to be specified to a file. Changes to keyword data are not even
transmitted to server as the local sandbox maintains a copy which is
diff'd against for spurious changes (keywords).  Merging and updates
also do not generate conflicts with keywords in Subversion. The kwdiff algorithm we had submitted in the patch can enable all these features that Subversion supports. Its a pity that the patch was rejected by CVS maintainers. The algorithm defined in the patch could have been also used to reduce the footprint of a CVS server process when doing RCS diff/compare, that was also
not considered useful.

We have made the patch public, it can be downloaded from -

I am also cc'ing info-cvs mailing list, so that other CVS users wanting Subversion like behavior could
use/modify the patch.

Patrick Richardson wrote:
Follow-up Comment #17, patch #4573 (project cvs):

It is really unfortunate that this bug will not be fixed. There has been no response to the case I made two and a half months ago FOR fixing the bug. It absolutely blows my mind that the CVS maintainers will not even consider implementing this patch as a non-default option.

Since it looks like the CVS maintainers will never fix this problem, I

had to recommend to my company that other alternatives should be evaluated, and if a suitable alternative is found, we should switch. After a careful evaluation of several open-source and commercial alternatives, we have decided to switch to Subversion.

Yes - the Subversion team got it right. Keyword expansion is Subversion works as CVS would IF the proposed patch were incorporated.


Rahul Bhargava,
SCM Solutions
Pleasanton, CA

reply via email to

[Prev in Thread] Current Thread [Next in Thread]