Re: [Tinycc-devel] CVS sync to Landley's tree (fork healing)

[Rob Landley]

On Friday 11 May 2007 10:38 am, David A. Wheeler wrote:
> And should Fabrice Bellard have more time in the future to devote to tinycc,
> he can use a process that he finds convenient (and then Landley can decide
> what he wants to do in HIS fork, in time convenient to him).

Just FYI:

The reason it took so long for me to grab Fabrice's -E changes is that he 
checked them into CVS rather than posting them to the list.  Extracting 
changes from CVS is a pain in the ass, because there's no way to get a list 
of them.  Instead you have to look at each file and try to match things up.  
(If there's a better way to do this, I haven't found it.  In theory tailor 
can improve on this, in practice my recent effort to make a mercurial copy of 
the qemu cvs with tailor-0.9.27 would only grab the last 256 changes no 
matter what I hit it with.)

Each time Fabrice showed any interest in tcc, I stopped working on my fork for 
a few months, to stay out of his way.  Part of the reason for that was that 
manually cherry picking ongoing changes from CVS into mercurial is hard 
enough I have no interest in even trying it.

So if your update of CVS results in changes actually going into CVS on a 
regular basis, I will either ignore them or walk away and let the CVS tree 
become the master repository again.

Just FYI.

> Of the changes that are reported as necessary to compile gcc 2.95, 8 are in
> Landley's tree, and 2 are in active discussion (my alloca, as well as the
> stringify work).  

Sorry, this week's been nuts. :)

> The script will save in CVS a common historical version between the current
> CVS and Landley's fork, and then add each changeset separately (so that much
> of the history can be retained in CVS).

Fabrice already imported a series of changes I sent him:

http://lists.gnu.org/archive/html/tinycc-devel/2006-10/msg00122.html

I believe that synced up everything through changeset 413 in my repository 
(which is why there were only 2 commits over the next 4 months).

> The script will occasionally be re-run from then on to keep them in sync. It
> will work by noticing the deltas in Landley's fork, and re-applying them;
> that way, if Fabrice Bellard makes a change solely in the CVS later on it
> won't be accidentally "lost".      

You're welcome to post it to the list, and I'll import it into my tree. :)

> Unfortunately, CVS doesn't really have "changeset" information, because CVS
> stores changes at the file level.

Sucks, doesn't it?

> CVS doesn't handle file renames gracefully, and I have no intentions\ 
> of "solving" that.       

Continues to suck, doesn't it?

Rob