[gentoo-dev] Committing straight to stable

[gentoo-dev] Committing straight to stable

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 801 bytes --]

Since keywording policy seems to be being ignored again... Don't *ever*
commit new ebuild revisions straight to stable, even if you think it's a
trivial fix. There are plenty of things that could go wrong even with
simple patches -- for example, if you accidentally included some CVS Id:
lines in your patch, they'll get nuked when you do the commit. And, if
you commit straight to stable, you end up breaking arch rather than just
~arch.

The "all things must go through ~arch for a while first" rule is there
for a good reason. It's not something you can arbitrarily ignore because
you think you're not breaking anything...

-- 
Ciaran McCreesh : Gentoo Developer (Vim, Fluxbox, shell tools)
Mail            : ciaranm at gentoo.org
Web             : http://dev.gentoo.org/~ciaranm


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

Re: [gentoo-dev] Committing straight to stable

[Aaron Walker]

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

Ciaran McCreesh wrote:

> Don't *ever* commit new ebuild revisions straight to stable, even if you think it's a
> trivial fix. 

Indeed! I learned that lesson with bug 73072 :)

- --
Aaron Walker <ka0ttic@gentoo.org>
[ BSD | cron | forensics | shell-tools | commonbox | netmon | vim | web-apps ]
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCa/bMC3poscuANHARAovUAKDWcWptHSHnmSg/LowMvrGgr+NJWACgiV67
AC9f+HPVGY/mah/TtaJbmjk=
=ZTS2
-----END PGP SIGNATURE-----
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Committing straight to stable

[Paul Waring]

On 4/24/05, Ciaran McCreesh <ciaranm@gentoo.org> wrote:
> Since keywording policy seems to be being ignored again... Don't *ever*
> commit new ebuild revisions straight to stable, even if you think it's a
> trivial fix. There are plenty of things that could go wrong even with
> simple patches -- for example, if you accidentally included some CVS Id:
> lines in your patch, they'll get nuked when you do the commit. And, if
> you commit straight to stable, you end up breaking arch rather than just
> ~arch.

Why not have a three strike rule - anyone who commits something
straight to stable 3 times in a given period (say 6 months) has their
CVS access revoked.

Paul

-- 
Rogue Tory
www.roguetory.org.uk

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Committing straight to stable

[Jan Kundrát]

[-- Attachment #1: Type: text/plain, Size: 278 bytes --]

Paul Waring wrote:
> Why not have a three strike rule - anyone who commits something
> straight to stable 3 times in a given period (say 6 months) has their
> CVS access revoked.

Heh, comming from non-gentoo email address :-)

-jkt

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

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

Re: [gentoo-dev] Committing straight to stable

[Paul Waring]

On 4/24/05, Jan Kundrát <jkt@flaska.net> wrote:
> Heh, comming from non-gentoo email address :-)

I get just as annoyed if someone commits something straight to stable
and it breaks something on my system. :)

Paul

-- 
Rogue Tory
www.roguetory.org.uk

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Committing straight to stable

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 1892 bytes --]

On Sun, 24 Apr 2005 21:29:19 +0100 Paul Waring <pwaring@gmail.com>
wrote:
| On 4/24/05, Ciaran McCreesh <ciaranm@gentoo.org> wrote:
| > Since keywording policy seems to be being ignored again... Don't
| > *ever* commit new ebuild revisions straight to stable, even if you
| > think it's a trivial fix. There are plenty of things that could go
| > wrong even with simple patches -- for example, if you accidentally
| > included some CVS Id: lines in your patch, they'll get nuked when
| > you do the commit. And, if you commit straight to stable, you end up
| > breaking arch rather than just ~arch.
| 
| Why not have a three strike rule - anyone who commits something
| straight to stable 3 times in a given period (say 6 months) has their
| CVS access revoked.

Because no-one would enforce it. As it stands right now, you can
repeatedly break eutils.eclass, all of profiles, keywords on any package
you like or anything else you care to name, and nothing will happen, no
matter how many people complain. On the other hand, close a single
bugzilla bug as INVALID and you risk the wrath of our esteemed devrel
team when someone complains.

What I'd *like* to see is all new devs and any dev who has a history of
breaking things committing to a branch rather than the main tree, and
having their commits approved (merged) by their mentor / someone sane.
Plus, possibly, having main-branch commits to things like eutils
restricted to people who really should be touching it. But CVS branches
are pretty much unusable in this respect...

*shrug* not that that's going to happen. Last I heard from someone in
devrel, this kind of thing was apparently the QA herd's area, not
theirs. Riiiiiiiiight...

-- 
Ciaran McCreesh : Gentoo Developer (Vim, Fluxbox, shell tools)
Mail            : ciaranm at gentoo.org
Web             : http://dev.gentoo.org/~ciaranm


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

Re: [gentoo-dev] Committing straight to stable

[Francesco Riosa]

>What I'd *like* to see is all new devs and any dev who has a history of
>breaking things committing to a branch rather than the main tree, and
>having their commits approved (merged) by their mentor / someone sane.
>  
>
Also if who approves is _not_ the mentor / sane a 4 eyes check is always
a good thing (TM) it's the way kernel develop is going from years now,
right?
Is it feasible ? Obviously without make the developer life an hell !

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Committing straight to stable

[Diego 'Flameeyes' Pettenò]

[-- Attachment #1: Type: text/plain, Size: 1663 bytes --]

On Sunday 24 April 2005 23:08, Francesco Riosa wrote:
> Also if who approves is _not_ the mentor / sane a 4 eyes check is always
> a good thing (TM) it's the way kernel develop is going from years now,
> right?
I think that doing something like that, surely will increase safety, but will 
also drive gentoo out of the world.

We have already too many packages which needs maintainers, and having to 
double-check every commit can be very very slow, because if there's too few 
people doing the second check, the bottleneck will stop everything being 
fixed, changed, updated.

Another problem is that there are tons of commits everyday, some of them are 
just trivials. I'm a new developer, but still today i did at least 10 
commits, if I counted them all. Some of them was just ~amd64 markings, other 
were fixes and version bumps. Some of them can't be tested, would require 
mergers to try the change locally and that could be really long, as to test 
some of them I needed to rebuild at least 6 packages. Also when keywording is 
concerned, after some time seeing similar patches you just can't say the 
differences between them.

I've done a couple of errors in this week I worked on gentoo, yes, but I was 
able to fix them asap. They was mainly trivial errors which I really have 
overseen (a ! not separed by space, an $Id: $ in a patch); probably 
double-checking them could have fixed them, but like so applying a security 
patch would have took surely more than just the about 40 minutes it took.

-- 
Diego "Flameeyes" Pettenò
Gentoo Developer (Gentoo/FreeBSD, Video, Gentoo/AMD64)

http://dev.gentoo.org/~flameeyes/

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

Re: [gentoo-dev] Committing straight to stable

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 1090 bytes --]

On Sun, 24 Apr 2005 23:24:08 +0200 "Diego 'Flameeyes' Pettenò"
<flameeyes@gentoo.org> wrote:
| I think that doing something like that, surely will increase safety,
| but will  also drive gentoo out of the world.
| 
| We have already too many packages which needs maintainers, and having
| to  double-check every commit can be very very slow, because if
| there's too few  people doing the second check, the bottleneck will
| stop everything being  fixed, changed, updated.

If it were implemented with CVS, then yes, because it'd be unusably
slow. With SVN, doing a branch merge for approval is extremely fast. And
your mentor is already checking all your commits, right? They're
supposed to do that for the first month or so (depending upon how long
it is before it becomes obvious that you're safe). A branch is just the
same, but without the added hazard of having commits going straight to
the live tree.

-- 
Ciaran McCreesh : Gentoo Developer (Vim, Fluxbox, shell tools)
Mail            : ciaranm at gentoo.org
Web             : http://dev.gentoo.org/~ciaranm


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

Re: [gentoo-dev] Committing straight to stable

[Diego 'Flameeyes' Pettenò]

[-- Attachment #1: Type: text/plain, Size: 430 bytes --]

On Sunday 24 April 2005 23:35, Ciaran McCreesh wrote:
> They're
> supposed to do that for the first month or so (depending upon how long
> it is before it becomes obvious that you're safe).
I was talking about double-checking *every* commit of every developer. That 
will be an overkill, imho.

-- 
Diego "Flameeyes" Pettenò
Gentoo Developer (Gentoo/FreeBSD, Video, Gentoo/AMD64)

http://dev.gentoo.org/~flameeyes/


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

Re: [gentoo-dev] Committing straight to stable

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 794 bytes --]

On Sun, 24 Apr 2005 23:37:00 +0200 "Diego 'Flameeyes' Pettenò"
<flameeyes@gentoo.org> wrote:
| On Sunday 24 April 2005 23:35, Ciaran McCreesh wrote:
| > They're
| > supposed to do that for the first month or so (depending upon how
| > long it is before it becomes obvious that you're safe).
|
| I was talking about double-checking *every* commit of every developer.
| That  will be an overkill, imho.

Oh, yeah, every developer would be a waste. But new developers and those
who have a history of breaking things (plus maybe most developers for
stuff like eutils) wouldn't be unreasonable if we had the technology...

-- 
Ciaran McCreesh : Gentoo Developer (Vim, Fluxbox, shell tools)
Mail            : ciaranm at gentoo.org
Web             : http://dev.gentoo.org/~ciaranm


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

Re: [gentoo-dev] Committing straight to stable

[Alec Warner]

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

Diego 'Flameeyes' Pettenò wrote:
> On Sunday 24 April 2005 23:35, Ciaran McCreesh wrote:
> 
>>They're
>>supposed to do that for the first month or so (depending upon how long
>>it is before it becomes obvious that you're safe).
> 
> I was talking about double-checking *every* commit of every developer. That 
> will be an overkill, imho.
> 

I think if you limited it to commits that are essentially tree
wide/eclasses it could be limited that way.  Fex, this weekend there was
a category move where repoman had some messed up behavior during the
move causing mrness headaches with recommiting a bunch of crap.  Things
like this as well as eclass changes ( which have the potential to break
100's of ebuilds/packages ) would be decent coverage.  As it stands
right now someone can make a tree-wide commit and fsck crap up and the
only recourse you have is to contact Infra and hope someone turns the
cvs->rsync off in time.  I think thats a glorious hack ( no offense
brian, hehe ) of an otherwise obvious ( to me anyway ) problem.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
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=XOel
-----END PGP SIGNATURE-----
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Committing straight to stable

[foser]

[-- Attachment #1: Type: text/plain, Size: 643 bytes --]

On Sun, 2005-04-24 at 21:29 +0100, Paul Waring wrote:
> Why not have a three strike rule - anyone who commits something
> straight to stable 3 times in a given period (say 6 months) has their
> CVS access revoked.

It's not California here. You completely ignore the fact that some
people commit more than others and as such are more likely to trip over
such a rule anyway and the people who do commit a lot are usually the
same people you don't want to revoke the access from.

It's just a reminder, don't make up some unenforceable policy just
because people make mistakes from time to time.

1 strike, you are out.

- foser

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

Re: [gentoo-dev] Committing straight to stable

[Paul Waring]

On 4/24/05, foser <foser@gentoo.org> wrote:
> It's not California here. You completely ignore the fact that some
> people commit more than others and as such are more likely to trip over
> such a rule anyway and the people who do commit a lot are usually the
> same people you don't want to revoke the access from.

It was only a suggestion - you could have it so that anyone who trips
up iin more than X% of commits gets access revoked, or reduced, or
whatever.
 
> 1 strike, you are out.

Oh I see, a user can't make one suggestion as to a policy to deal with
a problem but a dev can make as many mistakes as they like which cause
systems to break.

Paul

-- 
Rogue Tory
www.roguetory.org.uk

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Committing straight to stable

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 1065 bytes --]

On Sun, 24 Apr 2005 22:59:14 +0200 foser <foser@gentoo.org> wrote:
| On Sun, 2005-04-24 at 21:29 +0100, Paul Waring wrote:
| > Why not have a three strike rule - anyone who commits something
| > straight to stable 3 times in a given period (say 6 months) has
| > their CVS access revoked.
| 
| It's not California here. You completely ignore the fact that some
| people commit more than others and as such are more likely to trip
| over such a rule anyway and the people who do commit a lot are usually
| the same people you don't want to revoke the access from.

Actually, *most* of the regular high volume committers tend to get it
right, since they're experienced enough to know exactly what they're
doing and get everything right without even needing to think about it.
Of course, there's the occasional notable exception who regularly screws
stuff up and just plain doesn't care.

-- 
Ciaran McCreesh : Gentoo Developer (Vim, Fluxbox, shell tools)
Mail            : ciaranm at gentoo.org
Web             : http://dev.gentoo.org/~ciaranm


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

Re: [gentoo-dev] Committing straight to stable

[Mike Frysinger]

On Sunday 24 April 2005 05:09 pm, Ciaran McCreesh wrote:
> Of course, there's the occasional notable exception who regularly screws
> stuff up and just plain doesn't care.

spank me ! :(
-mike
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Committing straight to stable

[Anders Rune Jensen]

[-- Attachment #1: Type: text/plain, Size: 1307 bytes --]

On Sun, 2005-04-24 at 14:44 +0100, Ciaran McCreesh wrote:
> Since keywording policy seems to be being ignored again... Don't *ever*
> commit new ebuild revisions straight to stable, even if you think it's a
> trivial fix. There are plenty of things that could go wrong even with
> simple patches -- for example, if you accidentally included some CVS Id:
> lines in your patch, they'll get nuked when you do the commit. And, if
> you commit straight to stable, you end up breaking arch rather than just
> ~arch.
> 
> The "all things must go through ~arch for a while first" rule is there
> for a good reason. It's not something you can arbitrarily ignore because
> you think you're not breaking anything...

Let me first start by saying that committing straight to stable was
clearly a mistake. I can't help wonder why CVS would change patch files
(it probably doesn't know the difference between ordinary files and
patches) or why repoman doesn't catch something like this? CVS changing
files on commit goes against the whole "test before commit" mantra and
I'm probably not the first to have encountered this problem?

-- 
Anders Rune Jensen
http://www.cs.auc.dk/~arj/

PGP/GnuPG key: 1024D/62C2D7F0 @ pgp.mit.edu
Fingerprint: 6A03 907E 92E1 47EB 4EAB  76B6 068A ACD1 62C2 D7F0


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

Re: [gentoo-dev] Committing straight to stable

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1409 bytes --]

On Mon, Apr 25, 2005 at 03:04:48AM +0200, Anders Rune Jensen wrote:
> Let me first start by saying that committing straight to stable was
> clearly a mistake. I can't help wonder why CVS would change patch files
> (it probably doesn't know the difference between ordinary files and
> patches) 
This is mostly an assumption here, as I don't know all of the details,
but in the past I did run into problems with patches that dealt with
the CVS keywords. You'd commit the patch, but doing so would change it,
until you commited it with the -kb keyword option.

> or why repoman doesn't catch something like this? CVS changing
> files on commit goes against the whole "test before commit" mantra and
> I'm probably not the first to have encountered this problem?
You aren't the first no, but the solutions to it are limited:
- teach developers to use -kb where they should
- use an alternative keyword only for the gentoo-x86 module (some of
  the BSDs do this IIRC) - and turn off the normal keywords.
- repoman checks might be very difficult to do, but it should be
  possible to at least have it do warnings if it finds CVS keywords
  that might be dangerous.

-- 
Robin Hugh Johnson
E-Mail     : robbat2@orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 241 bytes --]

Re: [gentoo-dev] Committing straight to stable

[Donnie Berkholz]

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

Robin H. Johnson wrote:
> You aren't the first no, but the solutions to it are limited:
> - teach developers to use -kb where they should

Wouldn't it be -ko for a patch?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFCbFBKXVaO67S1rtsRAk/JAJ9NVf6yQWqAeEbIQAENRoT9xMAihgCfeRYd
SfsRmWO8JuzC2HzY3LR94Kk=
=T1Uo
-----END PGP SIGNATURE-----
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Committing straight to stable

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1818 bytes --]

On Sun, Apr 24, 2005 at 07:04:58PM -0700, Donnie Berkholz wrote:
> Robin H. Johnson wrote:
> > You aren't the first no, but the solutions to it are limited:
> > - teach developers to use -kb where they should
> Wouldn't it be -ko for a patch?
-kb is actually better than -ko, due to how it's handled on later commits to a file.

The CVS manual states:
`-ko' 
Generate the old keyword string, present in the working file just before it was
checked in. For example, for the Revision keyword, generate the string
$Revision: 1.1 $ instead of $Revision: 5.7 $ if that is how the string appeared
when the file was checked in. 
`-kb' 
Like `-ko', but also inhibit conversion of line endings between the canonical
form in which they are stored in the repository (linefeed only), and the form
appropriate to the operating system in use on the client. For systems, like
unix, which use linefeed only to terminate lines, this is very similar to
`-ko'. For more information on binary files, see Handling binary files. In CVS
version 1.12.2 and later `-kb', as set by cvs add, cvs admin, or cvs import may
not be overridden by a `-k' option specified on the command line. 

Note the last sentence again:
In CVS version 1.12.2 and later `-kb', as set by cvs add, cvs admin, or cvs
import may not be overridden by a `-k' option specified on the command line.

So this makes it much harder to break on later revisions as compared to -ko.

Also, if you have inconsistant linefeeds, cvs -ko can break sometimes (some
versions of Outlook mangled patches in this way @%$&@%$^).

-- 
Robin Hugh Johnson
E-Mail     : robbat2@orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 241 bytes --]

Re: [gentoo-dev] Committing straight to stable

[Donnie Berkholz]

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

Robin H. Johnson wrote:
> On Sun, Apr 24, 2005 at 07:04:58PM -0700, Donnie Berkholz wrote:
> 
>>Robin H. Johnson wrote:
>>
>>>You aren't the first no, but the solutions to it are limited:
>>>- teach developers to use -kb where they should
>>
>>Wouldn't it be -ko for a patch?
> 
> -kb is actually better than -ko, due to how it's handled on later commits to a file.

Interesting. Thanks for showing this to us -- I checked the info and man
pages but didn't catch this section.

Thanks,
Donnie
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCbG44XVaO67S1rtsRAgM6AJ9BdqfhxjC7RxWGP5DtSH61A1ErewCg6pb0
7gNP2ffJMbU23UPrLYOQxZg=
=TkiD
-----END PGP SIGNATURE-----
-- 
gentoo-dev@gentoo.org mailing list