Re: [PATCH] subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

[Yuya Nishihara]

On Mon, 16 Oct 2017 22:40:21 -0400, Matt Harbison wrote:
> On Mon, 16 Oct 2017 10:36:22 -0400, Yuya Nishihara  wrote:
> > On Mon, 16 Oct 2017 09:21:24 -0400, Matt Harbison wrote:
> >> +shared = hg.share(self._repo._subparent.baseui,
> >> +  other, self._repo.root,
> >> +  update=False, bookmarks=False)
> >
> > Perhaps bookmarks option should follow the parentrepo's, but that can be
> > addressed later.
> 
> I wondered about that.  But since the subrepo update may be deferred, and  
> I don't think the -B for the original share command is preserved, I wasn't  
> sure how.  (I don't think we can infer that -B was used, just because  
> there are bookmarks, right?)

It's stored in .hg/shared file.
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

Re: [PATCH] subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

[Matt Harbison]

On Tue, 17 Oct 2017 00:20:38 -0400, Matt Harbison   
wrote:


On Mon, 16 Oct 2017 10:00:32 -0400, Pulkit Goyal <7895pul...@gmail.com>  
wrote:


On Mon, Oct 16, 2017 at 6:51 PM, Matt Harbison   
wrote:

# HG changeset patch
# User Matt Harbison 
# Date 1508122082 14400
#  Sun Oct 15 22:48:02 2017 -0400
# Node ID d0c2b68fedb27184337af6392ecc1f03dab39522
# Parent  718adb1bf3a9a1ee509a803c7512c14296a1db79
subrepo: share instead of clone if the parent repo is shared  
(issue5675) (BC)


Can you please add a releasenotes part to the commit message?
For reference: https://www.mercurial-scm.org/wiki/ReleasenotesExtension


I took a run at it (probably not successfully- the command aborts, but  
check-code didn't complain).


$ hg --config extensions.releasenotes= releasenotes -r . rn.txt
abort: unexpected block in release notes directive fix


This went away after putting the 'and' line between the two annotations  
like Augie did in one of his commits.  That's pretty subtle- it would be  
nice if that requirement went away somehow.

___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

Re: [PATCH] subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

[Matt Harbison]

On Mon, 16 Oct 2017 10:00:32 -0400, Pulkit Goyal <7895pul...@gmail.com>  
wrote:


On Mon, Oct 16, 2017 at 6:51 PM, Matt Harbison   
wrote:

# HG changeset patch
# User Matt Harbison 
# Date 1508122082 14400
#  Sun Oct 15 22:48:02 2017 -0400
# Node ID d0c2b68fedb27184337af6392ecc1f03dab39522
# Parent  718adb1bf3a9a1ee509a803c7512c14296a1db79
subrepo: share instead of clone if the parent repo is shared  
(issue5675) (BC)


Can you please add a releasenotes part to the commit message?
For reference: https://www.mercurial-scm.org/wiki/ReleasenotesExtension


I took a run at it (probably not successfully- the command aborts, but  
check-code didn't complain).


$ hg --config extensions.releasenotes= releasenotes -r . rn.txt
abort: unexpected block in release notes directive fix

Is there any way to get the location of the error?  (I know nothing about  
this format, so I'm not sure where to start.)



In trying to reproduce this on a revision that's not mine (and in fact  
doesn't have an annotation), I got this:


$ hg --config extensions.releasenotes= releasenotes -r cfb054a7ecc4 rn.txt
** Unknown exception encountered with possibly-broken third-party  
extension evolve

** which supports versions 4.2 of Mercurial.
** Please disable evolve and try your action again.
** If that fixes the bug please report it to https://bz.mercurial-scm.org/
** Python 2.7.13 (v2.7.13:a06454b1afa1, Dec 17 2016, 20:53:40) [MSC v.1500  
64 bit (AMD64)]

** Mercurial Distributed SCM (version 4.3-rc+4-ac3477dc1814)
** Extensions loaded: eol, rebase, convert, graphlog, patchbomb, extdiff,  
strip, mq, evolve, releasenotes

Traceback (most recent call last):
  File "hg", line 45, in 
  File "mercurial\dispatch.pyc", line 81, in run
  File "mercurial\dispatch.pyc", line 162, in dispatch
  File "mercurial\dispatch.pyc", line 302, in _runcatch
  File "mercurial\dispatch.pyc", line 310, in _callcatch
  File "mercurial\scmutil.pyc", line 150, in callcatch
  File "mercurial\dispatch.pyc", line 292, in _runcatchfunc
  File "mercurial\dispatch.pyc", line 895, in _dispatch
  File "mercurial\dispatch.pyc", line 658, in runcommand
  File "mercurial\dispatch.pyc", line 903, in _runcommand
  File "mercurial\dispatch.pyc", line 892, in 
  File "mercurial\util.pyc", line 1069, in check
  File "mercurial\extensions.pyc", line 322, in closure
  File "mercurial\util.pyc", line 1069, in check
  File "hgext\mq.pyc", line 3542, in mqcommand
  File "mercurial\util.pyc", line 1069, in check
  File "hgext\releasenotes.pyc", line 440, in releasenotes
  File "hgext\releasenotes.pyc", line 339, in serializenotes
IndexError: list index out of range
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

Re: [PATCH] subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

[Matt Harbison]

On Mon, 16 Oct 2017 10:36:22 -0400, Yuya Nishihara  wrote:


On Mon, 16 Oct 2017 09:21:24 -0400, Matt Harbison wrote:

# HG changeset patch
# User Matt Harbison 
# Date 1508122082 14400
#  Sun Oct 15 22:48:02 2017 -0400
# Node ID d0c2b68fedb27184337af6392ecc1f03dab39522
# Parent  718adb1bf3a9a1ee509a803c7512c14296a1db79
subrepo: share instead of clone if the parent repo is shared  
(issue5675) (BC)


Generally looks good to me, but a few questions.


--- a/mercurial/hg.py
+++ b/mercurial/hg.py
@@ -255,6 +255,7 @@
 r = repository(ui, destwvfs.base)
 postshare(srcrepo, r, bookmarks=bookmarks, defaultpath=defaultpath)
 _postshareupdate(r, update, checkout=checkout)
+return r

 def postshare(sourcerepo, destrepo, bookmarks=True, defaultpath=None):
 """Called after a new shared repo is created.
diff --git a/mercurial/subrepo.py b/mercurial/subrepo.py
--- a/mercurial/subrepo.py
+++ b/mercurial/subrepo.py
@@ -857,28 +857,40 @@

 def _get(self, state):
 source, revision, kind = state
+parentrepo = self._repo._subparent
+
 if revision in self._repo.unfiltered():
-return True
+# Allow shared subrepos tracked at null to setup the  
sharedpath

+if revision != node.nullhex or not parentrepo.shared():


'revisions != node.nullhex' looks a bit tricky, which seemed to assume a
certain sequence how the current subrepo clone works under the hood.

Can we check 'len(self._repo) != 0' instead?


It looks like it.


+return True
 self._repo._subsource = source
 srcurl = _abssource(self._repo)
 other = hg.peer(self._repo, {}, srcurl)
 if len(self._repo) == 0:
-self.ui.status(_('cloning subrepo %s from %s\n')
-   % (subrelpath(self), srcurl))
-parentrepo = self._repo._subparent
 # use self._repo.vfs instead of self.wvfs to remove .hg  
only

 self._repo.vfs.rmtree()
-other, cloned = hg.clone(self._repo._subparent.baseui, {},
- other, self._repo.root,
- update=False)
-self._repo = cloned.local()
+if parentrepo.shared():
+self.ui.status(_('sharing subrepo %s from %s\n')
+   % (subrelpath(self), srcurl))
+shared = hg.share(self._repo._subparent.baseui,
+  other, self._repo.root,
+  update=False, bookmarks=False)


Perhaps bookmarks option should follow the parentrepo's, but that can be
addressed later.


I wondered about that.  But since the subrepo update may be deferred, and  
I don't think the -B for the original share command is preserved, I wasn't  
sure how.  (I don't think we can infer that -B was used, just because  
there are bookmarks, right?)



+self._repo = shared.local()
+else:
+self.ui.status(_('cloning subrepo %s from %s\n')
+   % (subrelpath(self), srcurl))
+other, cloned = hg.clone(self._repo._subparent.baseui,  
{},

+ other, self._repo.root,
+ update=False)
+self._repo = cloned.local()
 self._initrepo(parentrepo, source, create=True)
 self._cachestorehash(srcurl)
 else:
-self.ui.status(_('pulling subrepo %s from %s\n')
-   % (subrelpath(self), srcurl))
 cleansub = self.storeclean(srcurl)
-exchange.pull(self._repo, other)
+if not parentrepo.shared():


self._repo.shared() seems better since self._repo could be an existing
unshared repo.


I dropped this whole section.  Originally I was thinking that you never  
pull on update with a share.  But that aspect should be handled by the  
'revision in self._repo' check at the top.  You can `hg -R share pull  
some_other_clone`, so I think the pull code needs to stay.  (Though I  
failed to adapt the test to show this making a difference.)



+self.ui.status(_('pulling subrepo %s from %s\n')
+   % (subrelpath(self), srcurl))
+exchange.pull(self._repo, other)

___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

Re: [PATCH] subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

[Augie Fackler]

On Mon, Oct 16, 2017 at 07:30:32PM +0530, Pulkit Goyal wrote:
> On Mon, Oct 16, 2017 at 6:51 PM, Matt Harbison  wrote:
> > # HG changeset patch
> > # User Matt Harbison 
> > # Date 1508122082 14400
> > #  Sun Oct 15 22:48:02 2017 -0400
> > # Node ID d0c2b68fedb27184337af6392ecc1f03dab39522
> > # Parent  718adb1bf3a9a1ee509a803c7512c14296a1db79
> > subrepo: share instead of clone if the parent repo is shared (issue5675) 
> > (BC)
>
> Can you please add a releasenotes part to the commit message?
> For reference: https://www.mercurial-scm.org/wiki/ReleasenotesExtension

+1 - that'll make my life much easier on Wednesday!

> ___
> Mercurial-devel mailing list
> Mercurial-devel@mercurial-scm.org
> https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

Re: [PATCH] subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

[Yuya Nishihara]

On Mon, 16 Oct 2017 09:21:24 -0400, Matt Harbison wrote:
> # HG changeset patch
> # User Matt Harbison 
> # Date 1508122082 14400
> #  Sun Oct 15 22:48:02 2017 -0400
> # Node ID d0c2b68fedb27184337af6392ecc1f03dab39522
> # Parent  718adb1bf3a9a1ee509a803c7512c14296a1db79
> subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

Generally looks good to me, but a few questions.

> --- a/mercurial/hg.py
> +++ b/mercurial/hg.py
> @@ -255,6 +255,7 @@
>  r = repository(ui, destwvfs.base)
>  postshare(srcrepo, r, bookmarks=bookmarks, defaultpath=defaultpath)
>  _postshareupdate(r, update, checkout=checkout)
> +return r
>  
>  def postshare(sourcerepo, destrepo, bookmarks=True, defaultpath=None):
>  """Called after a new shared repo is created.
> diff --git a/mercurial/subrepo.py b/mercurial/subrepo.py
> --- a/mercurial/subrepo.py
> +++ b/mercurial/subrepo.py
> @@ -857,28 +857,40 @@
>  
>  def _get(self, state):
>  source, revision, kind = state
> +parentrepo = self._repo._subparent
> +
>  if revision in self._repo.unfiltered():
> -return True
> +# Allow shared subrepos tracked at null to setup the sharedpath
> +if revision != node.nullhex or not parentrepo.shared():

'revisions != node.nullhex' looks a bit tricky, which seemed to assume a
certain sequence how the current subrepo clone works under the hood.

Can we check 'len(self._repo) != 0' instead?

> +return True
>  self._repo._subsource = source
>  srcurl = _abssource(self._repo)
>  other = hg.peer(self._repo, {}, srcurl)
>  if len(self._repo) == 0:
> -self.ui.status(_('cloning subrepo %s from %s\n')
> -   % (subrelpath(self), srcurl))
> -parentrepo = self._repo._subparent
>  # use self._repo.vfs instead of self.wvfs to remove .hg only
>  self._repo.vfs.rmtree()
> -other, cloned = hg.clone(self._repo._subparent.baseui, {},
> - other, self._repo.root,
> - update=False)
> -self._repo = cloned.local()
> +if parentrepo.shared():
> +self.ui.status(_('sharing subrepo %s from %s\n')
> +   % (subrelpath(self), srcurl))
> +shared = hg.share(self._repo._subparent.baseui,
> +  other, self._repo.root,
> +  update=False, bookmarks=False)

Perhaps bookmarks option should follow the parentrepo's, but that can be
addressed later.

> +self._repo = shared.local()
> +else:
> +self.ui.status(_('cloning subrepo %s from %s\n')
> +   % (subrelpath(self), srcurl))
> +other, cloned = hg.clone(self._repo._subparent.baseui, {},
> + other, self._repo.root,
> + update=False)
> +self._repo = cloned.local()
>  self._initrepo(parentrepo, source, create=True)
>  self._cachestorehash(srcurl)
>  else:
> -self.ui.status(_('pulling subrepo %s from %s\n')
> -   % (subrelpath(self), srcurl))
>  cleansub = self.storeclean(srcurl)
> -exchange.pull(self._repo, other)
> +if not parentrepo.shared():

self._repo.shared() seems better since self._repo could be an existing
unshared repo.

> +self.ui.status(_('pulling subrepo %s from %s\n')
> +   % (subrelpath(self), srcurl))
> +exchange.pull(self._repo, other)
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel

Re: [PATCH] subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

[Pulkit Goyal]

On Mon, Oct 16, 2017 at 6:51 PM, Matt Harbison  wrote:
> # HG changeset patch
> # User Matt Harbison 
> # Date 1508122082 14400
> #  Sun Oct 15 22:48:02 2017 -0400
> # Node ID d0c2b68fedb27184337af6392ecc1f03dab39522
> # Parent  718adb1bf3a9a1ee509a803c7512c14296a1db79
> subrepo: share instead of clone if the parent repo is shared (issue5675) (BC)

Can you please add a releasenotes part to the commit message?
For reference: https://www.mercurial-scm.org/wiki/ReleasenotesExtension
___
Mercurial-devel mailing list
Mercurial-devel@mercurial-scm.org
https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel