[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [cross-project-issues-dev] CVS access to Orbit
|
I hate breaking things -- I really do. The HTTP protocol is beautiful
in that respect -- there are built-in provisions for redirecting old
links to new ones. I could easily do something similar with pserver
cvs, where the server would respond with a plaintext "CVS is replaced
with Git" message that could be thrown at the user as an unexpected
server response. However, such a message would break CVS access for Orbit.
Maintaining pserver for all our old repos is possible -- after all,
we've been doing it for 13 years! But cvs is arguably no longer
maintained, it has a habit of not always exiting cleanly leaving zombie
processes, its archaic file locking is prone to errors and it is Yet
Another Service your webmaster team would need to actively maintain.
We've added Git and Gerrit was simply too cool to pass up, so CVS just
had to go.
Beyond that, we have a steady stream of university researchers who pull
in our code and go on analyzing it for years. For those users, and for
potential new contributors, it has to be clear that code in CVS is not
an accurate portrait of the current state at Eclipse. We agreed back in
2008 that having three SCM's at Eclipse didn't make sense. We've
stretched the migration timeline as much as we could. CVS has done good
by us and now it is gone. I suggest we now focus our energies on
updating those references to CVS to point to new resources as we stumble
upon them.
Thanks,
Denis
On 01/15/2013 11:19 AM, Ed Willink wrote:
Hi
December 21st was the end of the world as far as committer CVS write
access was concerned.
However it was not the end of the world as far as user CVS read access
was concerned.
A variety of examples and documentation describe fetching code from
CVS. It is not feasible
to rewrite such text within Helios/Indigo/Juno distributions taht are
still in use.
It is also a major pain for us committers. For instance, when a
Bugzilla report refers to line 45 of version 1.9 of a file;
without viewcvs, it takes significantly longer to identify the line in
question.
There seems absolutely no benefit in retracting this useful facility
and breaking functionality upon which
the current Juno release documentation depends.
Regards
Ed Willink
On 15/01/2013 15:14, Doug Schaefer wrote:
While Denis is pealing himself off the ceiling, I'll chime in. Isn't
this
all in git now? I thought Orbit was the only thing officially left in
CVS.
:D
On 13-01-15 10:08 AM, "Ed Willink"<ed@xxxxxxxxxxxxx> wrote:
Hi
/cvsroot/modeling etc etc
Regards
Ed Willink
On 15/01/2013 14:58, Denis Roy wrote:
What other repositories? The only CVS repo I know of is Orbit.
Denis
On 01/15/2013 03:36 AM, Ed Willink wrote:
Hi Denis
Thanks. CVS from PSFs works again.
If we now have an anonymous CVS server for Orbit, can we please
restore the anonymous CVS access for the other repositories too so
that Helios, Indigo, Juno facilities work again for ordinary users?
Regards
Ed Willink
On 14/01/2013 22:29, Denis Roy wrote:
I've re-enabled pserver CVS for /cvsroot/tools. I'm hoping someone
can try it out, since my modern-day computer doesn't ship with cvs
:-D
Denis
On 01/14/2013 03:36 PM, Ed Willink wrote:
Hi
My understanding is that CVS has been preserved for Orbit, at least
temporarily, but anonymous access has not.
How am I expected to rewrite PSF fetches (for use by ordinary
users) such as
<provider id="org.eclipse.team.cvs.core.cvsnature">
<project
reference="1.0,:pserver:anonymous@xxxxxxxxxxxxxxx:/cvsroot/tools,org.e
clipse.orbit/lpg.runtime.java,lpg.runtime.java,v2_0_17"/>
</provider>
Regards
Ed Willink
