RE: [Axiom-developer] Axiom silver branch

[Bill Page]

On April 10, 2006 12:06 PM Ralf Hemmecke wrote:
>
> Bill Page wrote:
> > In the SourceForge Axiom CVS archive I have set the binary mode
> > for all the *.tgz files in the zips directory using the command:
> > 
> > cvs admin -kb
> > 
> > as described here:
> > 
> > http://www.idevelopment.info/data/Programming/change_managemen
> >   t/unix_cvs/PROGRAMMING_Working_with_Binary_Files.shtml

> Hmm, I guess, if the binary flag for the file in CVS was not put
> there BEFORE the file is first checked into CVS then a later change
> will not help. I have not tested that, but I think that the file
> is already corrupted.

No. There is no corruption in the CVS. You can do a checkout of
Axiom from the CVS and everything works fine, even though the
binary flag is not set. CVS actually assumes unix line endings
so there is really no conversion of the binary provided that you
only use CVS on a linux/unix platform. A problem only occurs with
CVS if you use a Windows (or other OS) version of CVS where the
line ending defaults to something else. Even in that case with
the newer versions of Windows one can simply specify that you
want to keep the "unix-style" line endings and everything works
out fine.

The problem occurs when you transfer a CVS containing binary
files that are not properly marked as such, to a SVN repository.
In that case the files are corrupted even on linux/unix. I do
not know the reason why this occurs in SVN - whether it is a
bug or just a design "feature". Anyway, but setting the correct
flags on CVS and *then* doing another transfer of the contents
of the CVS to SVN, everything should work out ok. (probably
not the last word ... ;)

> Why cannot we simply use the current patch-47 (plus Tim's
> changes) and generate the SVN stuff from that?

As far as I understand it, patch-47 is the content of the CVS
on SourceForge.

>
> I think it would even be reasonable to release patch-48 and
> then start with these sources the SVN archive. Then we can
> forget about all the trouble with CVS and binaries.

As I said, the trouble starts with SVN, not CVS.

> Does someone need the history that is in the CVS currently?
> Aren't we satisfied with the history of axiom--main--1--patch-47?

I think you might be right about the irrelevance of history
to the "silver" branch so that means we could optimize the
SVN archive and remove (most?) of this history.

Regards,
Bill Page.