bzr branch fetches way too much data with stacked branches

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Aaron Bentley wrote:
> Public bug reported:
>
> I pushed the same branch to launchpad with branch and push, and the
> results were dramatically different:
>
> $ bzr branch . :push
> created bzr+ssh://~abentley/launchpad/bmp-to-addr/
> Its packs direrectory listing:
> rw-r--r-- 184M 373e5989a9fde59a76ab9788cfd0b757.pack
>
> $ bzr push bzr+ssh://bazaar.launchpad.net/~abentley/launchpad/bmp-to-addr-pushed
> created bzr+ssh://bazaar.launchpad.net/%7Eabentley/launchpad/bmp-to-addr-push/
> Its packs directory listing
> rw-r--r-- 2.1M 16cb335d253769d5a91504342d819077.pack
> rw-r--r-- 3.3M 98765e55aaa396d3fc4114e198db0d42.pack
>
> Both branches are stacked on ~launchpad-pqm/launchpad/db-devel.

I would guess that "bzr branch . lp..." isn't respecting the stacked
flag at all, and is just pushing up the whole history.

That looks approximately right given the sizes you list.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkpVMboACgkQJdeBCYSNAAPRpwCeMuO9wpJ/VvkXqBTM0Dj3OhNc
F64An1VK51cQLwNJ9GAX65H1o7Z5Aigg
=hCDW
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

John A Meinel wrote:
> Aaron Bentley wrote:
>> Public bug reported:
>
>> I pushed the same branch to launchpad with branch and push, and the
>> results were dramatically different:
>
>> $ bzr branch . :push
>> created bzr+ssh://~abentley/launchpad/bmp-to-addr/
>> Its packs direrectory listing:
>> rw-r--r-- 184M 373e5989a9fde59a76ab9788cfd0b757.pack
>
>> $ bzr push bzr+ssh://bazaar.launchpad.net/~abentley/launchpad/bmp-to-addr-pushed
>> created bzr+ssh://bazaar.launchpad.net/%7Eabentley/launchpad/bmp-to-addr-push/
>> Its packs directory listing
>> rw-r--r-- 2.1M 16cb335d253769d5a91504342d819077.pack
>> rw-r--r-- 3.3M 98765e55aaa396d3fc4114e198db0d42.pack
>
>> Both branches are stacked on ~launchpad-pqm/launchpad/db-devel.
>
> I would guess that "bzr branch . lp..." isn't respecting the stacked
> flag at all, and is just pushing up the whole history.

It is reflecting the stacked flag some:
$ bzr branch . :push
Using default stacking branch /~launchpad-pqm/launchpad/db-devel at
bzr+ssh://bazaar.launchpad.net/%7Eabentley/launchpad/
Created new stacked branch referring to /~launchpad-pqm/launchpad/db-devel.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkpVNQ4ACgkQ0F+nu1YWqI0tLQCfZBAJFS+Rcac3FlOPz6V7i3Xd
qaUAn3fBIYPTYFC/tTf5hG9+l8ydQ+sb
=cnTI
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Aaron Bentley wrote:
> John A Meinel wrote:
>> Aaron Bentley wrote:
>>> Public bug reported:
>>> I pushed the same branch to launchpad with branch and push, and the
>>> results were dramatically different:
>>> $ bzr branch . :push
>>> created bzr+ssh://~abentley/launchpad/bmp-to-addr/
>>> Its packs direrectory listing:
>>> rw-r--r-- 184M 373e5989a9fde59a76ab9788cfd0b757.pack
>>> $ bzr push bzr+ssh://bazaar.launchpad.net/~abentley/launchpad/bmp-to-addr-pushed
>>> created bzr+ssh://bazaar.launchpad.net/%7Eabentley/launchpad/bmp-to-addr-push/
>>> Its packs directory listing
>>> rw-r--r-- 2.1M 16cb335d253769d5a91504342d819077.pack
>>> rw-r--r-- 3.3M 98765e55aaa396d3fc4114e198db0d42.pack
>>> Both branches are stacked on ~launchpad-pqm/launchpad/db-devel.
>> I would guess that "bzr branch . lp..." isn't respecting the stacked
>> flag at all, and is just pushing up the whole history.
>
> It is reflecting the stacked flag some:
> $ bzr branch . :push
> Using default stacking branch /~launchpad-pqm/launchpad/db-devel at
> bzr+ssh://bazaar.launchpad.net/%7Eabentley/launchpad/
> Created new stacked branch referring to /~launchpad-pqm/launchpad/db-devel.
>
> Aaron

Sure. It could, for example, set the 'stacked-on-url' in the Branch
config, but not actually set a fallback repository on its Repository
object. I'm certainly just guessing, but I think it is a reasonable
guess of where to start looking.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkpVTNEACgkQJdeBCYSNAANoAgCeMceg+8fh788LiFvPVEDPEYDQ
94QAoIMbX7A4G2ovrQnPUn7L2MWX3B8M
=Hfob
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

John A Meinel wrote:
>>> flag at all, and is just pushing up the whole history.
>> It is reflecting the stacked flag some:
>> $ bzr branch . :push
>> Using default stacking branch /~launchpad-pqm/launchpad/db-devel at
>> bzr+ssh://bazaar.launchpad.net/%7Eabentley/launchpad/
>> Created new stacked branch referring to /~launchpad-pqm/launchpad/db-devel.

> Sure. It could, for example, set the 'stacked-on-url' in the Branch
> config, but not actually set a fallback repository on its Repository
> object. I'm certainly just guessing, but I think it is a reasonable
> guess of where to start looking.

AFAICT, it affects clone_on_transport also, but only if revision_id is
None. (Perhaps push forces the revision_id higher up the chain and
branch does not.)

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkpVeoUACgkQ0F+nu1YWqI0PNgCfSuS5F2UUhPpZoeDv35LSG9ZB
epUAn2Uim31XxqG8Rfpbc3Jsl+QUCpzp
=ErVf
-----END PGP SIGNATURE-----