default plugin path
robertc at robertcollins.net
Tue Jan 27 03:32:17 GMT 2009
On Mon, 2009-01-26 at 22:12 -0500, Aaron Bentley wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> Robert Collins wrote:
> >>> which is a dead chicken.
> >> If you mean it's voodoo, I disagree. It's reasonably easy to understand.
> > Its easy to understand but its replicating logic that in the common case
> > isn't needed.
> There is no common case.
Sure there is - pqm, every microscript I've seen, bzr itself, all want
the bzrlib default path setting logic. AFAIK only launchpad and
bundlebuggy want something different.
> >>> What harmful effect will setting the path have?
> >> It will cause bzrlib clients to load bzr plugins from the system path,
> >> which will be incompatible with their local version of bzrlib.
> > If I put a patch forward that doesn't spuriously add the system python
> > to the path
> Adding the system python to the path is *right* for many cases. It's
> also *wrong* for many cases. Therefore the client should decide.
Alternatively, which is what I'm proposing, the clients that consider it
wrong can just change it. Their code looks identical in both cases. So
its transparent to them.
> > (there is a bug in the path setting at the moment affecting
> > normal bzr as well as bzr clients), what harmful effects would setting
> > the path by default at module load have?
> It would be setting bzrlib to load plugins that the client author hasn't
No. Its telling bzrlib *where* plugins are found, but in itself isn't
One of the following applies:
- the client author doesn't load plugins at all - non issue
- the client author wants the same plugins bzr loads - no action needed
vis-a-vis the path
- the client author wants a specific path - they set the path as
> >> Launchpad and Bundle Buggy have both been hit by bad default plugin
> >> paths, even though Launchpad tried to control the plugin path.
> > I know that - but I'm not proposing to *load plugins*, only to set the
> > path.
> Which would mean that the wrong plugins would be loaded if we used the
> "from foo import bar" method.
The same logic is used for 'from bzrlib.plugins.foo import bar' as for
'load_plugins()' - a search path is searched to find foo, and it is
imported into bzrlib.plugins.foo.
> > And both can set it to whatever they like *anyway*.
> Yeah, but they need to remember to do that. These are the sort of bad
> defaults that look right up until the point that stuff starts breaking
> for all users.
There are currently two bits of damage:
- the path is wrong for *all users* including bzr itself. bug
https://bugs.edge.launchpad.net/bzr/+bug/316192 has discussion on this.
- clients have to know to call a function to allow callouts to *any*
plugin to work.
I'm happy to make the second depend on fixing the first, to avoid
breaking clients like lp that were affected by bug 316192. But I don't
think that the situation today is better than what we had in e.g. bzrlib
1.0! And the simplest comparison is that more code is needed now than
then - clearly a regression IMO.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 197 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20090127/42735c4a/attachment.pgp
More information about the bazaar