Re: Time to start wrapping up the conversion.

[Ethan A M. <sf...@us...>]

[trying again]
My mailer ate the embedded traceback.


On Wednesday, November 1, 2017 2:22:43 PM PDT Eric S. Raymond wrote:
> I've uploaded a fresh version of the conversion machinery where you
> can fetch it at
> 
> wget http://www.catb.org/~esr/gnuplot-conversion.tar.gz

For me, reconvert dIes before finishing:


reposurgeon% # The 4.0 branch 
reposurgeon% blob <FUBAR 
reposurgeon% /#include gp_time.h/ mailbox_in --create <NEWCOMMIT 
reposurgeon% /#include gp_time.h/ assign v40root --singleton 
Traceback (most recent call last): 
 File "/usr/local/bin/reposurgeon", line 12393, in <module> 
   main() 
 File "/usr/local/bin/reposurgeon", line 12369, in main 
   interactive() 
 File "/usr/local/bin/reposurgeon", line 12358, in interactive 
   interpreter.cmdloop() 
 File "/usr/lib64/python2.7/cmd.py", line 141, in cmdloop 
   line = self.precmd(line) 
 File "/usr/local/bin/reposurgeon", line 7960, in precmd 
   line = self.set_selection_set(line) 
 File "/usr/local/bin/reposurgeon", line 7996, in set_selection_set 
   if self.chosen().named(self.line): 
 File "/usr/local/bin/reposurgeon", line 5750, in named 
   key=len, reverse=True): # longest name first 
TypeError: object of type 'NoneType' has no len()


Suggestions?


> This version contains a new script, 'forcepush'. After running
> reconvert, you can run forcepush to nuke the the SourceForge git
> repository and replace it with the new conversion.  (You must
> have Tcl and Expect installed for this to work.)
> 
> The script does assume that your local username is the sane as your
> SourceForge one, because I'm 'esr' both places.  If this is a
> problem for anyone, I'm sure we can modify the "nuke" and "forcepush"
> scripts to generalize them.
> The entire conversion and replacement process is now implemented and
> documented by the three scripts reconvert, nuke, and forcepush.  It
> can be run at any time by anyone with the right SourceForge
> credentials.
> 
> When doing repo conversions, I think it's important to show the work
> so all its assumptions are auditable.  Also, we are now far enough
> along in the process that if I have to run off to focus exclusively on
> something else, somebody (most likely Dan Sebald or HBB) would be able
> to finish the job.  

Sounds good.

> So, this is my major deliverable; the rest is polishing.

Thanks. Really, thanks.


> Things still to be done:
> 
> 1. The EMPTIES file in the conversion tarball needs to have real log
> messages filled in for all those "*** empty log message ***" instances
> (but don't touch the Check-Text headers; those are a safety measure
> in case the commit dates drift, they should prevent clobbering of the
> comment text in nearby commits).
>
> A relatively easy way to do this: bring up EMPTIES in your favorite
> editor, then run gitk on a repo conversion.  Looking at the diffs
> gitk displays per commit should make it easy to write a useful change
> comment.  Note that the topmost text box in the lower half of the
> gitk display is for text search; if you type "*** empty log message
> ***" into it, you can use the up and down arrows off to the left
> to quickly move between empty comments.

Not going to happen.  If I had nothing to say about a commit at the
time, I certainly don't have anything to say about it now.
Is there a command to replace all of these with the Jedi mind trick 
"This is not the commit you are looking for"
 
> If all of you mail your EMPTIES changes back to me I'll take care of
> collating them.  If the process runs long I might set up a
> public repo for the conversion stuff.
> 
> 2. The final disposition of the 3.7.x branch needs to be decided.  I
> think it is soaking up more effort than it is worth and should be
> dropped.  I'm pretty sure Ethan agrees.

Yes.

> We're almost done.  Let's get this wrapped up.

!!
	Ethan