Re: Feature request: confirm data loss when renaming folder with ignored files

Greg Troxel <[email protected]>
Newsgroups gmane.network.unison.general
Message-ID <[email protected]>
Tõivo Leedjärv <[email protected]> writes:

>> I think the answer is probably more or less "because ignored files are
>> not important, by the definition of what should be ignored."
>
> I'm a bit torn about this statement. On one hand, "ignore" means
> "pretend it's not there" (for purpose of the sync) and in that sense
> Unison has not directly done anything wrong. One the other hand,
> "ignore" does not mean "not important" and it should not mean "go
> ahead and delete these for me (or alter in some way)". Some kind of
> warning in the UI might make sense.

An explanation not aspirational.  And I don't think warn is good enough;
I think we need "just won't".


> It's not exactly new knowledge that propagating renames the way it's
> done now can under specific conditions lead to temporary data loss, in
> some cases even to permanent data loss, and I've also submitted bug
> fixes in that spirit.

yes, but the permanent loss is a conceptual bug, not a consequence of a
crash  in the middle.[q
>
> A slight change of topic. Thinking about the strict view of "never
> alter ignored files/dirs in any way", I looked at preferences like
> "nodeletionpartial" and "noupdatepartial". A "cautious" option, as
> suggested by Jorge, could automatically produce "nodeletionpartial" or
> "noupdatepartial" preferences based on "ignore" preferences. However,
> I just discovered that the "nodeletionpartial" preference does not
> work as expected because it affects the update reconciliation step,
> but not the propagation step. As such, a -nodeletionpartial 'Path
> sub/subsub -> root1' will happily allow deleting the entire path
> 'sub', including 'sub/subsub' in it.

I can't even follow this.  That's more about me, being a bit tired this
minute.  But things are already too complicated.

> I suppose the entire topic of "don't touch what you're not supposed to
> touch" requires some rethinking when it comes to deleting (parent)
> directories (this includes renaming). I don't intend to do it myself
> but it would be interesting to hear ideas from people interested in
> this topic.

My take is:

  do not do any recursive deletes, ever

  only delete a directory if it is empty.  Otherwise, error out and make
  the user fix it

  hope this is better, and learn what the trouble is

  do something else later, maybe
  

-- 
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.