Launchpad svn/git imports failure causes

[Alexander Belchenko]

So, I've just tried to execute

bzr get lp:sikuli

Bazaar downloaded more than 130MB (one hundred and thirty MB)of data
from Internet and then just refused to create a working tree. No
wonder there is unhappy users.

I've looked at help for view command. It lacks the --exclude option,
IMO. How can one define the view which cover entire tree without
couple of files? That's just crazy. And from the description of views
it does not seem it will help us because views applied to
status/diff/commit commands, but you need to create a working tree
first, and this operation is failed.

As I understand, svn, git and hg just create text placeholder file
instead of real symlink. Can we implement this for bzr in easy way? I
don't know is it really possible. My old win32symlinks plugin last
time I've tried it was failed to handle all cases, because the code to
check whether the path is symlink or regular file seems to be spilled
through several modules in bzrlib.

Russel Winder пишет:
> Martin,
> 
> On Fri, 2010-09-10 at 17:06 +0100, Martin (gzlist) wrote:
> [ . . . ]
>> This approach of treating some platforms as second-class and
>> discouraging non-portable features doesn't really seem to be working
>> for symlinks. Pretty much always the right answer is "don't version
>> symlinks if you claim to be a cross-platform project", but instead we
>> get things like <https://bugs.launchpad.net/sikuli/+bug/586709> which
>> gets turned into a dupe of our bug 81689 rather than getting that
>> straight forward advice, then this cute little follow up from antatoly
>> <https://bugs.launchpad.net/sikuli/+bug/634791> which could be
>> paraphrased as "switch to mercurial".
> 
> I fear some of this may be my fault.  The context is the SCons developer
> and user lists.  I am trying to foment the use of DVCS for add-on SCons
> tools -- traditionally all tools were in the core distribution or posted
> on a wiki.  I really like Bazaar/Launchpad for these and so have been
> using them (cf. http://www.scons.org/wiki/ToolsIndex), though
> Git/GitHub, Mercurial/BitBucket are other possibilities.  However I
> really like the fact that Launchpad supports teams as well as
> individuals, whereas GitHub and BitBucket are all about individuals.
> 
> Anatoly, in response to one of my emails, raised the issue of the Sikuli
> problem https://answers.edge.launchpad.net/sikuli/+question/100764 and
> implied that no Ubuntu developers cared about Windows -- the point is
> nothing to do with SCons or my email, but he spotted I was somewhat
> knowledgeable about Bazaar, I have pushed Bazaar previously on the SCons
> lists, though the decision has now been made that SCons core will switch
> from Subversion to Mercurial.  I berated him a bit about over-inferring,
> i.e. no Bazaar developer would have seen the question, and that he
> should make a posting associated with Bazaar so that the right people
> saw the problem and so could investigate.  Clearly
> https://bugs.launchpad.net/sikuli/+bug/634791 was his response to that,
> which actually annoys me somewhat as it doesn't address the right
> problem, and isn't aimed at the right audience.
> 
> Anyway, the fact that there is an issue is now a topic in the right
> audience.  Do we have any idea what the actual problem is that Anatoly
> is seeing with Sikuli?  Is it simply that the committers have assumed
> symlink capability which Windows fails to have?  Is there a lesson here
> for Bazaar?  Can Mercurial really get round these issues or is Anatoly's
> suggestion just a "peeved hack" that may not solve the problem?
>