[gentoo-dev] RFC: removal of digest files from the tree

[gentoo-dev] RFC: removal of digest files from the tree

[Zac Medico]

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

Now that the whole tree has been converted to Manifest2 format we can remove the
old digest files from the tree so that it is pure Manifest2 format [1]. Any
users who still have a version of portage that doesn't support Manifest2 (less
than sys-apps/portage-2.1 or stages prior to 2006.1) will still be able to use a
script to generate the old digests if they want to upgrade an old system [2].

In order to trigger removal of the old digests, all that we have to do is add a
file named manifest1_obsolete to the root of the repository. Presence of that
file causes repoman to automatically remove the old digests upon commit. If
people don't update their whole cvs tree then they are likely to miss this file
which will cause them to generate the old digest files and add them back to the
tree unnecessarily. If that happens then it's not a major problem but we should
try to make sure that everybody gets the file in order to prevent confusion.

After we've added the manifest1_obsolete file we can either let the old digest
files disappear gradually or we can actively remove them.  It's possible to use
repoman to remove them from a single package, from a whole category, or from the
whole tree at once.  The repoman --force option can be used to make it work
faster by skipping dependency checks. If we want to actively remove them then
the best approach is probably to use repoman to remove them from one category at
a time.

If there are no objections then I don't so any reason not to go ahead and add
the manifest1_obsolete sometime in the near future. Thoughts?

Zac

[1]
http://www.gentoo.org/proj/en/glep/glep-0044.html#removal-of-digest-files
[2] http://www.gentoo.org/proj/en/portage/doc/converting-manifest2.xml

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)

iD8DBQFHm9mY/ejvha5XGaMRAgFYAKCKKTZjYbLl1E/Djj/5VQp2jMzT9QCeK6N/
Ze0SmeWEhOsF4J3avthW/TQ=
=/HZV
-----END PGP SIGNATURE-----
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 2221 bytes --]

On Sat, Jan 26, 2008 at 05:08:42PM -0800, Zac Medico wrote:
> In order to trigger removal of the old digests, all that we have to do is add a
> file named manifest1_obsolete to the root of the repository. Presence of that
> file causes repoman to automatically remove the old digests upon commit. If
> people don't update their whole cvs tree then they are likely to miss this file
> which will cause them to generate the old digest files and add them back to the
> tree unnecessarily. If that happens then it's not a major problem but we should
> try to make sure that everybody gets the file in order to prevent confusion.
Related to the top level of the tree, can we get a release of repoman
that detects if a checkout is a subset only (eg no top level), for folk
that have subtree checkouts only? (Probably look for skel.*/profile
items two levels up).

> After we've added the manifest1_obsolete file we can either let the old digest
> files disappear gradually or we can actively remove them.  It's possible to use
> repoman to remove them from a single package, from a whole category, or from the
> whole tree at once.  The repoman --force option can be used to make it work
> faster by skipping dependency checks. If we want to actively remove them then
> the best approach is probably to use repoman to remove them from one category at
> a time.
I vote for actively removing them. I can do big commits like that on the
CVS server, and save the network turnaround time that would happen
otherwise.

> If there are no objections then I don't so any reason not to go ahead and add
> the manifest1_obsolete sometime in the near future. Thoughts?
Let's do it. I look forward to a lot less inodes on my disks.

Doing some rough calculations:
~6700/12400 packages have only digests in their files/ directory.
CVS overhead removal from pruned files/ directories:
6700*5 = 33500 inodes (3 files in CVS/, plus 2 directories).
Removal of direct digest-$PV files:
25335 files, 1 inode per file.
=====
58835 inodes total
=====

-- 
Robin Hugh Johnson
Gentoo Linux Developer & Infra Guy
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 329 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Josh Saddler]

[-- Attachment #1: Type: text/plain, Size: 552 bytes --]

Zac Medico wrote:
> (shtuff)

This sounds like something that might require extensive documentation
changes. If so, please review the Portage Handbooks for anything that
will need updating, as well as the Gentoo Upgrading Guide[1]. This
latter document is where we keep the instructions for upgrading from old
installations. Not sure if this removal will seriously affect such users
or not. If so, patches and/or additional instructions sent to the GDP
will be most welcome. Thanks!

[1] http://www.gentoo.org/doc/en/gentoo-upgrading.xml


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Rémi Cardona]

Robin H. Johnson a écrit :
> Related to the top level of the tree, can we get a release of repoman
> that detects if a checkout is a subset only (eg no top level), for folk
> that have subtree checkouts only? (Probably look for skel.*/profile
> items two levels up).

If repoman needs to check the existence of ../../manifest1_obsolete, 
people with subtree checkouts will still be committing digests right?

Shouldn't we hard code into repoman that generating digests shouldn't be 
done anymore, whether the manifest1_obsolete file exists or not? (making 
that file obsolete in the process :) )

> 58835 inodes total

/me eagerly waits for CVS updates to be _much_ faster.

Rémi
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Fabian Groffen]

On 26-01-2008 17:35:37 -0800, Robin H. Johnson wrote:
> Related to the top level of the tree, can we get a release of repoman
> that detects if a checkout is a subset only (eg no top level), for folk
> that have subtree checkouts only? (Probably look for skel.*/profile
> items two levels up).

Is it possible to have a CVS server-side commit check to see if digest
files were added, and if so to reject the commit?


-- 
Fabian Groffen
Gentoo on a different level
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Roy Bamford]

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

On 2008.01.27 01:08, Zac Medico wrote:
> Now that the whole tree has been converted to Manifest2 format we can
> remove the old digest files from the tree so that it is pure 
> Manifest2 format [1]. Any
> users who still have a version of portage that doesn't support
> Manifest2 (less than sys-apps/portage-2.1 or stages prior to 2006.1) 
> will still be able to use a script to generate the old digests if 
> they want to upgrade an old system [2].
>
[snip]
> If there are no objections then I don't so any reason not to go ahead
> and add the manifest1_obsolete sometime in the near future. Thoughts?
> 
> Zac
> 
> [1]
> http://www.gentoo.org/proj/en/glep/glep-0044.html#removal-of-digest-
> files
> [2] http://www.gentoo.org/proj/en/portage/doc/converting-
> manifest2.xml
> 
> 

Team,

We need to get news items out, on gentoo.org, in the GWN and on the 
forums or everyone with an old portage will be unhappy. Some users will 
still be unhappy anyway but we can point to the news and say we tried 
to tell you. If link [2] will be around for a while, we can us it in an 
mzbot factoid too.

No objections ... just plenty of publicity.

- -- 
Regards,

Roy Bamford
(NeddySeagoon) a member of
gentoo-ops
forum-mods
treecleaners

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)

iD8DBQFHnICfTE4/y7nJvasRAjWlAKCowpUt9i1CpdmqD7UyB6DXxzgOcQCg7+k0
t4RRbBHMuGSOhySOhmLV2II=
=PvBz
-----END PGP SIGNATURE-----

--
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Petteri Räty]

[-- Attachment #1: Type: text/plain, Size: 542 bytes --]

Fabian Groffen kirjoitti:
> On 26-01-2008 17:35:37 -0800, Robin H. Johnson wrote:
>> Related to the top level of the tree, can we get a release of repoman
>> that detects if a checkout is a subset only (eg no top level), for folk
>> that have subtree checkouts only? (Probably look for skel.*/profile
>> items two levels up).
> 
> Is it possible to have a CVS server-side commit check to see if digest
> files were added, and if so to reject the commit?
> 

Should be as we already fail invalid .xml files.

Regards,
Petteri


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1068 bytes --]

On Sun, Jan 27, 2008 at 04:06:47PM +0200, Petteri R??ty wrote:
> Fabian Groffen kirjoitti:
>> On 26-01-2008 17:35:37 -0800, Robin H. Johnson wrote:
>>> Related to the top level of the tree, can we get a release of repoman
>>> that detects if a checkout is a subset only (eg no top level), for folk
>>> that have subtree checkouts only? (Probably look for skel.*/profile
>>> items two levels up).
>> Is it possible to have a CVS server-side commit check to see if digest
>> files were added, and if so to reject the commit?
> Should be as we already fail invalid .xml files.
Due to how CVS hooks operate, it's not quite possible.
You wouldn't be able to block the entire commit, only the contents of
the files/ directory would get totally blocked.
If you were committing an ebuild along with a patch, this would be very
bad, as the ebuild+Manifest would get committed, but the patch wouldn't.

-- 
Robin Hugh Johnson
Gentoo Linux Developer & Infra Guy
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 329 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Petteri Räty]

[-- Attachment #1: Type: text/plain, Size: 1069 bytes --]

Robin H. Johnson kirjoitti:
> On Sun, Jan 27, 2008 at 04:06:47PM +0200, Petteri R??ty wrote:
>> Fabian Groffen kirjoitti:
>>> On 26-01-2008 17:35:37 -0800, Robin H. Johnson wrote:
>>>> Related to the top level of the tree, can we get a release of repoman
>>>> that detects if a checkout is a subset only (eg no top level), for folk
>>>> that have subtree checkouts only? (Probably look for skel.*/profile
>>>> items two levels up).
>>> Is it possible to have a CVS server-side commit check to see if digest
>>> files were added, and if so to reject the commit?
>> Should be as we already fail invalid .xml files.
> Due to how CVS hooks operate, it's not quite possible.
> You wouldn't be able to block the entire commit, only the contents of
> the files/ directory would get totally blocked.
> If you were committing an ebuild along with a patch, this would be very
> bad, as the ebuild+Manifest would get committed, but the patch wouldn't.
> 

Then perhaps we could set the hook to send a mail to the developer in 
question.

Regards,
Petteri


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Zac Medico]

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

Robin H. Johnson wrote:
> Related to the top level of the tree, can we get a release of repoman
> that detects if a checkout is a subset only (eg no top level), for folk
> that have subtree checkouts only? (Probably look for skel.*/profile
> items two levels up).

I think we should do like Rémi suggested and make it unconditionally
behave as if the manifest1_obsolete file exists. Once we've dropped
the old digest files from the repo, the behavior doesn't need to be
conditional anymore.

> I vote for actively removing them. I can do big commits like that on the
> CVS server, and save the network turnaround time that would happen
> otherwise.
> 
>> If there are no objections then I don't so any reason not to go ahead and add
>> the manifest1_obsolete sometime in the near future. Thoughts?
> Let's do it. I look forward to a lot less inodes on my disks.

Let's schedule a date for it then and we can have a big announcement
to let everyone know.

Zac
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)

iD8DBQFHnSu2/ejvha5XGaMRAjhgAJ9/2dXtokv0jG4kuRPzn8AFC0IAiwCgu5SF
HD9X3MqWZ0WilMDl9HbpuXU=
=sNXP
-----END PGP SIGNATURE-----
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 614 bytes --]

On Sun, 2008-01-27 at 06:27 -0800, Robin H. Johnson wrote:
> Due to how CVS hooks operate, it's not quite possible.
> You wouldn't be able to block the entire commit, only the contents of
> the files/ directory would get totally blocked.
> If you were committing an ebuild along with a patch, this would be very
> bad, as the ebuild+Manifest would get committed, but the patch wouldn't.

Bleh... CVS vs. SVN.  There's no "pre-commit" equivalent on CVS?

Also, wouldn't the second Manifest run "fix" the missing digest commit?

-- 
Chris Gianelloni
Release Engineering Strategic Lead
Games Developer

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1483 bytes --]

On Sat, 2008-01-26 at 17:08 -0800, Zac Medico wrote:
> If there are no objections then I don't so any reason not to go ahead and add
> the manifest1_obsolete sometime in the near future. Thoughts?

1. Add blocking of commit of files/digest-* in CVS pre-commit hook
2. Add manifest1_obsolete to tree
3. Remove all files/digest-* files from tree (on CVS server side)
4. Remove all files dirs that are empty (on CVS server side)

This should be all that we need.  The 2007.1 snapshot was done without
any digest files, and it worked just fine, also the default settings
have been to not sync the digest files for some time, so only people
with an older/unsupported portage have been getting digest files, at
all.

I do have one question, though.  What does an older portage version do
when it hits a package with a missing digest file?

Let's say I've got portage prior to 2007.0's, so it doesn't support
Manifest2 only.  I want to "emerge --oneshot portage" to get to the
latest version.  What do I need a digest on?  Just portage?  portage and
its dependencies?  Which dependencies?  All of them, or just the ones
I'll actually need to merge?

I guess what I'm asking is if it is possible to have repoman create a
digest for sys-apps/portage only, for the slow upgraders.  Of course,
this assumes its even possible on the portage side to bother.  If not,
just ignore this part.

-- 
Chris Gianelloni
Release Engineering Strategic Lead
Games Developer

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1862 bytes --]

On Sun, 2008-01-27 at 17:11 -0800, Zac Medico wrote:
> >> If there are no objections then I don't so any reason not to go ahead and add
> >> the manifest1_obsolete sometime in the near future. Thoughts?
> > Let's do it. I look forward to a lot less inodes on my disks.
> 
> Let's schedule a date for it then and we can have a big announcement
> to let everyone know.

Dammit.

I wanted this before January 31st, so it would make it into the 2008.0
snapshot.  Do we think we can get this done by January 31st?

I think so.  One reason that I think that we don't need to wait very
long is rather simple.  People running very old versions of portage that
will be affected by this are also not likely to read our news or see our
front page on a regular basis.  They're likely to get broken, no matter
what we do.  Hell, even posting it to the GMN/forums/lists/planet/front
page, we'll still end up getting complaints from these people when they
come back $months from now and the news is no longer sticky in the
forums, has rotated off the front page, isn't even a distant memory on
the lists, and is in a several month old newsletter.

When gauging impact/scope of a problem, always look at who it affects
and the situation.  You only need to take as much precaution as
necessary to cover the cases worth spending the time to cover.  There
will *always* be corner cases.  You just try to minimize them.

I doubt much of anything aside from portage/paludis/pkgcore use the
digest files, and I'd bet that paludis/pkgcore don't use them, at all.

I think throwing up an announcement today/tomorrow for Thursday/Friday
should be sufficient for this sort of a change, as it won't affect any
user who has a version of portage released in the past ~1.5 years.

-- 
Chris Gianelloni
Release Engineering Strategic Lead
Games Developer

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1030 bytes --]

On Sun, Jan 27, 2008 at 05:34:14PM -0800, Chris Gianelloni wrote:
> On Sun, 2008-01-27 at 06:27 -0800, Robin H. Johnson wrote:
> > Due to how CVS hooks operate, it's not quite possible.
> > You wouldn't be able to block the entire commit, only the contents of
> > the files/ directory would get totally blocked.
> > If you were committing an ebuild along with a patch, this would be very
> > bad, as the ebuild+Manifest would get committed, but the patch wouldn't.
> Bleh... CVS vs. SVN.  There's no "pre-commit" equivalent on CVS?
There is pre-commit, but it's not recursive. It gets applied to only a
single directory and in isolation from the other directories.

> Also, wouldn't the second Manifest run "fix" the missing digest commit?
In what way? the problem I was concerned about was non-digest files
not being committed leading to broken ebuilds.

-- 
Robin Hugh Johnson
Gentoo Linux Developer & Infra Guy
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 329 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Zac Medico]

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

Chris Gianelloni wrote:
> I do have one question, though.  What does an older portage version do
> when it hits a package with a missing digest file?
> 
> Let's say I've got portage prior to 2007.0's, so it doesn't support
> Manifest2 only.  I want to "emerge --oneshot portage" to get to the
> latest version.  What do I need a digest on?  Just portage?  portage and
> its dependencies?  Which dependencies?  All of them, or just the ones
> I'll actually need to merge?
> 
> I guess what I'm asking is if it is possible to have repoman create a
> digest for sys-apps/portage only, for the slow upgraders.  Of course,
> this assumes its even possible on the portage side to bother.  If not,
> just ignore this part.
> 

All that they have to do is run the manifest221 script [1] and it
will generate the old style digest files containing SHA1 digests
from the Manifest files. I've tested that script on a 2005.0 stage3
and I was able to upgrade it pretty smoothly.

[1] http://www.gentoo.org/proj/en/portage/doc/converting-manifest2.xml

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)

iD8DBQFHnUIw/ejvha5XGaMRAi1XAJ9JE0aE9MwuSF+XRBi4SrMwFtkpxQCgpC2Z
WEHpxtkRaquXWwLV34Nv8LI=
=o11+
-----END PGP SIGNATURE-----
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1187 bytes --]

On Sun, 2008-01-27 at 18:01 -0800, Robin H. Johnson wrote:
> On Sun, Jan 27, 2008 at 05:34:14PM -0800, Chris Gianelloni wrote:
> > On Sun, 2008-01-27 at 06:27 -0800, Robin H. Johnson wrote:
> > > Due to how CVS hooks operate, it's not quite possible.
> > > You wouldn't be able to block the entire commit, only the contents of
> > > the files/ directory would get totally blocked.
> > > If you were committing an ebuild along with a patch, this would be very
> > > bad, as the ebuild+Manifest would get committed, but the patch wouldn't.
> > Bleh... CVS vs. SVN.  There's no "pre-commit" equivalent on CVS?
> There is pre-commit, but it's not recursive. It gets applied to only a
> single directory and in isolation from the other directories.

OK.  So we could block on commits of digest-* files to files, right?
What else would we need?

> > Also, wouldn't the second Manifest run "fix" the missing digest commit?
> In what way? the problem I was concerned about was non-digest files
> not being committed leading to broken ebuilds.

Ahh, never mind... I was thinking of something else.

-- 
Chris Gianelloni
Release Engineering Strategic Lead
Games Developer

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Vlastimil Babka]

Chris Gianelloni wrote:
> On Sun, 2008-01-27 at 18:01 -0800, Robin H. Johnson wrote:
>> On Sun, Jan 27, 2008 at 05:34:14PM -0800, Chris Gianelloni wrote:
>>> On Sun, 2008-01-27 at 06:27 -0800, Robin H. Johnson wrote:
>>>> Due to how CVS hooks operate, it's not quite possible.
>>>> You wouldn't be able to block the entire commit, only the contents of
>>>> the files/ directory would get totally blocked.
>>>> If you were committing an ebuild along with a patch, this would be very
>>>> bad, as the ebuild+Manifest would get committed, but the patch wouldn't.
>>> Bleh... CVS vs. SVN.  There's no "pre-commit" equivalent on CVS?
>> There is pre-commit, but it's not recursive. It gets applied to only a
>> single directory and in isolation from the other directories.
> 
> OK.  So we could block on commits of digest-* files to files, right?
> What else would we need?

IMHO even if it's possible to selectively block digests and let other 
FILESDIR stuff (patches) through, it won't remove unneeded 
files/digest-* entries in Manifest.

I think we shouldn't overreact to this relatively small problem. If few 
people commit few digests back, big deal, won't hurt anyone. As 
betelgeuse said, just make the hook send mail to the dev so he can fix it.

Caster

-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Vlastimil Babka]

Chris Gianelloni wrote:
> On Sun, 2008-01-27 at 17:11 -0800, Zac Medico wrote:
>>>> If there are no objections then I don't so any reason not to go ahead and add
>>>> the manifest1_obsolete sometime in the near future. Thoughts?
>>> Let's do it. I look forward to a lot less inodes on my disks.
>> Let's schedule a date for it then and we can have a big announcement
>> to let everyone know.
> 
> Dammit.
> 
> I wanted this before January 31st, so it would make it into the 2008.0
> snapshot.  Do we think we can get this done by January 31st?

Would be great.

> I think so.  One reason that I think that we don't need to wait very
> long is rather simple.  People running very old versions of portage that
> will be affected by this are also not likely to read our news or see our
> front page on a regular basis.  They're likely to get broken, no matter
> what we do.  Hell, even posting it to the GMN/forums/lists/planet/front
> page, we'll still end up getting complaints from these people when they
> come back $months from now and the news is no longer sticky in the
> forums, has rotated off the front page, isn't even a distant memory on
> the lists, and is in a several month old newsletter.

Sounds logical.

> I think throwing up an announcement today/tomorrow for Thursday/Friday
> should be sufficient for this sort of a change, as it won't affect any
> user who has a version of portage released in the past ~1.5 years.
> 

++

Caster
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Gilles Dartiguelongue]

Le dimanche 27 janvier 2008 à 17:44 -0800, Chris Gianelloni a écrit :
> On Sun, 2008-01-27 at 17:11 -0800, Zac Medico wrote:
> > >> If there are no objections then I don't so any reason not to go ahead and add
> > >> the manifest1_obsolete sometime in the near future. Thoughts?
> > > Let's do it. I look forward to a lot less inodes on my disks.
> > 
> > Let's schedule a date for it then and we can have a big announcement
> > to let everyone know.
> 
> Dammit.
> 
> I wanted this before January 31st, so it would make it into the 2008.0
> snapshot.  Do we think we can get this done by January 31st?
[snip]
> I think throwing up an announcement today/tomorrow for Thursday/Friday
> should be sufficient for this sort of a change, as it won't affect any
> user who has a version of portage released in the past ~1.5 years.

+1 from me
-- 
Gilles Dartiguelongue <eva@gentoo.org>
Gentoo

--
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Rémi Cardona]

Zac Medico a écrit :
> Robin H. Johnson wrote:
>> Related to the top level of the tree, can we get a release of repoman
>> that detects if a checkout is a subset only (eg no top level), for folk
>> that have subtree checkouts only? (Probably look for skel.*/profile
>> items two levels up).
> 
> I think we should do like Rémi suggested and make it unconditionally
> behave as if the manifest1_obsolete file exists. Once we've dropped
> the old digest files from the repo, the behavior doesn't need to be
> conditional anymore.

We can do both my suggestion and Chris's as well.

Let's schedule the "touch manifest1_obsolete" for Thursday/Friday, and 
we can release an updated version of portage/repoman that doesn't even 
know -digest files ever existed.

Best of both worlds :)

Rémi
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Santiago M. Mola]

On Jan 28, 2008 2:44 AM, Chris Gianelloni <wolf31o2@gentoo.org> wrote:
> I think throwing up an announcement today/tomorrow for Thursday/Friday
> should be sufficient for this sort of a change, as it won't affect any
> user who has a version of portage released in the past ~1.5 years.

+1, too.

-- 
Santiago M. Mola
Jabber ID: cooldwind@gmail.com
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Vlastimil Babka]

[-- Attachment #1: Type: text/plain, Size: 411 bytes --]

Chris Gianelloni wrote:
> I think throwing up an announcement today/tomorrow for Thursday/Friday
> should be sufficient for this sort of a change, as it won't affect any
> user who has a version of portage released in the past ~1.5 years.

So, since it seems nobody objected it, time for the announcement? And 
the removal (server-side by robbat2) would be best done right before the 
snapshot?

VB


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 751 bytes --]

On Thu, Jan 31, 2008 at 12:05:26AM +0100, Vlastimil Babka wrote:
> Chris Gianelloni wrote:
>> I think throwing up an announcement today/tomorrow for Thursday/Friday
>> should be sufficient for this sort of a change, as it won't affect any
>> user who has a version of portage released in the past ~1.5 years.
> So, since it seems nobody objected it, time for the announcement? And the 
> removal (server-side by robbat2) would be best done right before the 
> snapshot?
It's pretty much irrelevant to the snapshot. wolf31o2 has my patch to
exclude digests in the snapshot anyway.

-- 
Robin Hugh Johnson
Gentoo Linux Developer & Infra Guy
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 329 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Santiago M. Mola]

On Jan 31, 2008 12:32 AM, Robin H. Johnson <robbat2@gentoo.org> wrote:
>
> On Thu, Jan 31, 2008 at 12:05:26AM +0100, Vlastimil Babka wrote:
> > Chris Gianelloni wrote:
> >> I think throwing up an announcement today/tomorrow for Thursday/Friday
> >> should be sufficient for this sort of a change, as it won't affect any
> >> user who has a version of portage released in the past ~1.5 years.
> > So, since it seems nobody objected it, time for the announcement? And the
> > removal (server-side by robbat2) would be best done right before the
> > snapshot?
> It's pretty much irrelevant to the snapshot. wolf31o2 has my patch to
> exclude digests in the snapshot anyway.
>

Then I guess it could be done as soon as the doc patch is prepared...

-- 
Santiago M. Mola
Jabber ID: cooldwind@gmail.com
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Chris Gianelloni]

On Thu, 2008-01-31 at 00:45 +0100, Santiago M. Mola wrote:
> On Jan 31, 2008 12:32 AM, Robin H. Johnson <robbat2@gentoo.org> wrote:
> >
> > On Thu, Jan 31, 2008 at 12:05:26AM +0100, Vlastimil Babka wrote:
> > > Chris Gianelloni wrote:
> > >> I think throwing up an announcement today/tomorrow for Thursday/Friday
> > >> should be sufficient for this sort of a change, as it won't affect any
> > >> user who has a version of portage released in the past ~1.5 years.
> > > So, since it seems nobody objected it, time for the announcement? And the
> > > removal (server-side by robbat2) would be best done right before the
> > > snapshot?
> > It's pretty much irrelevant to the snapshot. wolf31o2 has my patch to
> > exclude digests in the snapshot anyway.
> >
> 
> Then I guess it could be done as soon as the doc patch is prepared...

It really is best done in the tree first, so everything is correct for
the snapshot.  We didn't have much issue last time because we push
"manifest1_obsolete" into the snapshot, so digests are removed, anyway,
but I'd *really* prefer if this were done prior to the snapshot being
made.

The less fudging I have to do with the snapshot, the better.

-- 
Chris Gianelloni
Release Engineering Strategic Lead
Games Developer
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 662 bytes --]

On Wed, Jan 30, 2008 at 04:14:48PM -0800, Chris Gianelloni wrote:
> It really is best done in the tree first, so everything is correct for
> the snapshot.  We didn't have much issue last time because we push
> "manifest1_obsolete" into the snapshot, so digests are removed, anyway,
> but I'd *really* prefer if this were done prior to the snapshot being
> made.
Ok, so we're just blocking on the docs guys.

Are they actively working on it so I can go ahead and take the digests
out anyway?

-- 
Robin Hugh Johnson
Gentoo Linux Developer & Infra Guy
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 329 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Davide Cendron]

[-- Attachment #1: Type: text/plain, Size: 1200 bytes --]

Il Thursday 31 January 2008 02:00:36 Robin H. Johnson ha scritto:
> On Wed, Jan 30, 2008 at 04:14:48PM -0800, Chris Gianelloni wrote:
> > It really is best done in the tree first, so everything is correct for
> > the snapshot.  We didn't have much issue last time because we push
> > "manifest1_obsolete" into the snapshot, so digests are removed, anyway,
> > but I'd *really* prefer if this were done prior to the snapshot being
> > made.
>
> Ok, so we're just blocking on the docs guys.
>
> Are they actively working on it so I can go ahead and take the digests
> out anyway?

jkt has recently committed some fix:

http://sources.gentoo.org/viewcvs.py/gentoo/xml/htdocs/doc/en/bugzilla-howto.xml?r1=1.10&r2=1.11
http://sources.gentoo.org/viewcvs.py/gentoo/xml/htdocs/doc/en/gentoo-ppc-faq.xml?r1=1.66&r2=1.67
http://sources.gentoo.org/viewcvs.py/gentoo/xml/htdocs/doc/en/handbook/hb-working-portage.xml?r1=1.68&r2=1.69

I don't know exactly how many other docs must be changed accordingly, it's 
better to ping some GDP member to get a better feedback.

Cheers,

-- 
Davide Cendron

Gentoo Documentation Project
Italian Lead Translator

http://www.gentoo.org/doc/it/

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Jan Kundrát]

[-- Attachment #1: Type: text/plain, Size: 1787 bytes --]

Davide Cendron wrote:
>> Ok, so we're just blocking on the docs guys.

Well, to the best of my knowledge, nobody asked us to fix stuff.
Preferred way would be filing a bug to the docs-team that says "hey,
we're gonna deprecate all digests, could you please fix your crap so
that it doesn't recommend running `ebuild foo digest`?", but other
methods like mailing the -dev ML or pinging us on irc are fine, too.

If you don't want to be held back by our inactivity, you have to let us
know :).

Also note that docs team is responsible for files that are under
http://www.gentoo.org/doc/ and *nothing else*. If there's a bug in some
other document that is on our web nodes, like for example the Devrel
Handbook, we leave it to their maintainers (devrel here) to fix it. If
such team says "hey, fell free to fix whatever you want", I'm sure some
GDP members (including rane myself) will gladly help, but right now, the
policy is "we got a bug that speaks about stuff under /proj/ -> reassign".

(As a side note, we have that great devmanual which seems a lot more
useful than the ebuild part of the devrel handbook, perhaps it's time to
reconsider their roles. And yeah, I've brought this to devrel's
attention and don't count on them following -dev :). )

> I don't know exactly how many other docs must be changed accordingly, it's 
> better to ping some GDP member to get a better feedback.

Quick grep suggests that these needs fixing:

proj/en/desktop/games/games-ebuild-HOWTO.txt
proj/en/devrel/handbook/hb-guide-ebuild.xml
proj/en/devrel/handbook/hb-introduction-new-devs.xml
proj/en/devrel/new-dev-training.xml

I'm sure that rane will  get someone from the devrel to fix them :)

Cheers,
-jkt

-- 
cd /local/pub && more beer > /dev/mouth


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: [gentoo-dev] RFC: removal of digest files from the tree

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 810 bytes --]

On Thu, Jan 31, 2008 at 10:57:01PM +0100, Jan Kundr?t wrote:
> Davide Cendron wrote:
> >> Ok, so we're just blocking on the docs guys.
> Well, to the best of my knowledge, nobody asked us to fix stuff.
...
> Also note that docs team is responsible for files that are under
> http://www.gentoo.org/doc/ and *nothing else*.
nightmorph was the one that suggested doc changes were needed in /doc/:
http://archives.gentoo.org/gentoo-dev/msg_40d4cdd191d19898e5231410a7a512b6.xml

zmedico gave this document pointer about what users with old systems
need to know:
http://archives.gentoo.org/gentoo-dev/msg_038fd120fd1150be8d64500edf9cd851.xml

-- 
Robin Hugh Johnson
Gentoo Linux Developer & Infra Guy
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 329 bytes --]