Re: [Gnu-arch-users] Re: baz bug: network-access despite local repo

[Ulf Ochsenfahrt]

On Sun, 2005-06-19 at 15:31 -0500, John Arbash Meinel wrote:
> Matthieu Moy wrote:
> 
> >Ulf Ochsenfahrt <address@hidden> writes:
> >
> > [ ... snip ... ]
> >
> My guess is that he branched from a remote repository, and "baz commit"
> is trying to generate the complete ancestry for the branch. That
> requires connecting to the old archive.

That is a bad idea. The ancestry (according to CONTINUATION) is:

address@hidden/cqs--mainline--0.1--patch-6
address@hidden/cqs--ulfjack--0.1--version-0
address@hidden/cqs--mainline--0.2--version-0
address@hidden/cqs--mainline--0.2

That old archive may not even exist anymore. (I'm actually wondering
that it is still registered.) What's worse is that it first commited the
patch to the archive, then went to look for the archive and DID NOT mark
the tree has having the patch.

I had to Ctrl-C it because I didn't have network at that time and when I
tried to continue with tla, tla told me that the tree was not
up-to-date. So i tla updated it and - guess what - I got a conflict.

Why would it go looking for the ancestry anyway?

> I think there was a bug at one point where when trying to create the
> ancestry, baz did not consider the fact that the ancestry was already
> generated at an earlier stage. (Say patch-50 has +ancestry.gz, but 51
> does not (generated by tla) it would be nice if 52 could realize that 50
> exists, and not have to do the rest of the back-tracking.)
> 
> John
> =:->

In order to better support disconnected operation, baz should _never_
silently do network operations when not explicitly asked to do them. I
only have a slow (yet expensive) dialup line and that is already a pain.
If the tools don't work properly without network, that is unacceptable.

Thanks for your help.

Cheers,

-- Ulf

signature.asc
Description: This is a digitally signed message part