Current-Users archive

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

Re: git copies of cvs modules available

At Wed, 28 Oct 2009 20:12:55 -0400, Arnaud Lacombe 
<> wrote:
Subject: Re: git copies of cvs modules available
> well, as an heavy git user, what I want to know is if the source I'm
> looking at are the same as the one who has an issue with the source.

Yup, it's the same old problem programmers have always had!  ;-)

Actually that _is_ what the revision number in a CVS/RCS/SCCS style
system with a central repository effectively does give to you.

No, there's no guarantees, but the assumptions are usually valid for
most cases.  (SCCS is strictly always better at this than RCS/CVS under
normal use, though RCS/CVS do have the option of "freezing" the ID
values in copies of released files such that it won't normally be
changed even if the user re-imports the released files into their own
separate RCS/CVS repository.)

In distributed systems such as Git, Hg, and so on the use of a hash
value is (apparently) necessary internally, but from a human point of
view this hash string really does not successfully replace that
human-(semi-)comprehensible revision number, not even for the most basic
use of one-to-one comparison and especially not for the case of guessing
about distance down the revision tree.

In the end the only true way to know if you're looking at the same
source or not is to get a verbatim copy of the actually source the
problem is being reported in and then to independently reproduce the
problem first from that copy of the source from the complainant, and
then again from what you believe to be identical source in your own
repository.  Sure it's easier to use some shorter representation of the
original source file (a hash string, or even that old SCCS/CVS/RCS
revision number), especially if what you distribute is actually a
compiled binary representation of the original source.  Even then unless
you can somehow reproduce the reported problem using a functionally
equivalent binary built from known sources, you're just guessing until
you get to the point where you can have the original complainant try a
fixed version in their own environment.

                                                Greg A. Woods
                                                Planix, Inc.

<>       +1 416 218 0099

Attachment: pgpSK6Vz66cJE.pgp
Description: PGP signature

Home | Main Index | Thread Index | Old Index