[gentoo-dev] File collisions

[gentoo-dev] File collisions

[Steve Long]

On the issue of QA, I think enabling FEATURES="collision-detect" by default
would do a lot more good at this stage than "test". There are several
packages I have come across where this has picked up files not being
removed from earlier versions. It's a reasonable default to have and
doesn't involve delaying users since it takes a fraction of a second to
carry out. What I'd propose is that portage runs the collision-detect and
reports it, but continues the installation in the same manner as now.

This could tie into an automated report at a later stage.

According to the portage team the majority of blockers in the tree are
`soft' blocks which are file collisions. I understand that they'd like to
separate those from hard blocks which are more fundamental
incompatibilities than installing the same file, since that would make it
easier for QA to track those. collision-detect isn't going to help with
that, of course, but I wonder what the issues would be with such a change?


-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] File collisions

[Ioannis Aslanidis]

On 4/19/07, Steve Long <slong@rathaus.eclipse.co.uk> wrote:
> On the issue of QA, I think enabling FEATURES="collision-detect" by default
> would do a lot more good at this stage than "test". There are several
> packages I have come across where this has picked up files not being
> removed from earlier versions. It's a reasonable default to have and
> doesn't involve delaying users since it takes a fraction of a second to
> carry out. What I'd propose is that portage runs the collision-detect and
> reports it, but continues the installation in the same manner as now.
>
> This could tie into an automated report at a later stage.

Do you mean to report a 'QA notice' instead of halting the merge?
Well, actually halting the merge is a required feature, so maybe the
QA notice should be added always straightway, not bound to the
FEATURE.

-- 
Ioannis Aslanidis

<deathwing00[at]gentoo.org> 0xB9B11F4E
-- 
gentoo-dev@gentoo.org mailing list

[gentoo-dev] Re: File collisions

[Steve Long]

Ioannis Aslanidis wrote:
> On 4/19/07, Steve Long wrote:
>> On the issue of QA, I think enabling FEATURES="collision-detect" by
>> default would do a lot more good at this stage than "test". There are
>> several packages I have come across where this has picked up files not
>> being removed from earlier versions. It's a reasonable default to have
>> and doesn't involve delaying users since it takes a fraction of a second
>> to carry out. What I'd propose is that portage runs the collision-detect
>> and reports it, but continues the installation in the same manner as now.
>>
>> This could tie into an automated report at a later stage.
> 
> Do you mean to report a 'QA notice' instead of halting the merge?
> Well, actually halting the merge is a required feature, so maybe the
> QA notice should be added always straightway, not bound to the
> FEATURE.
> 
Yes, that's what I meant, sorry for being unclear. The collision-detect
should run by default, and report a QA error, only halting if the FEATURE
is set.

-- 
gentoo-dev@gentoo.org mailing list

[gentoo-dev] Re: File collisions

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 308 bytes --]

Steve Long <slong@rathaus.eclipse.co.uk>:

> On the issue of QA, I think enabling FEATURES="collision-detect" by
> default would do a lot more good at this stage than "test".

 Arch teams normally have collision-protect enabled when doing
keywording/stabling....in my eyes this is sufficient.

V-Li

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

Re: [gentoo-dev] Re: File collisions

[Rob C]

[-- Attachment #1: Type: text/plain, Size: 546 bytes --]

On 19/04/07, Christian Faulhammer <opfer@gentoo.org> wrote:
>
> Steve Long <slong@rathaus.eclipse.co.uk>:
>
> > On the issue of QA, I think enabling FEATURES="collision-detect" by
> > default would do a lot more good at this stage than "test".
>
> Arch teams normally have collision-protect enabled when doing
> keywording/stabling....in my eyes this is sufficient.
>
> V-Li
>
>
Its obviously not, Many users are reporting file-collisions on a weekly
basis. So either this isn't sufficient or the arch teams are not acting as
you describe.

-Rob

[-- Attachment #2: Type: text/html, Size: 944 bytes --]

Re: [gentoo-dev] Re: File collisions

[Jakub Moc]

On 4/20/07, Rob C <hyakuhei@gentoo.org> wrote:
> On 19/04/07, Christian Faulhammer <opfer@gentoo.org> wrote:
> > Arch teams normally have collision-protect enabled when doing
> > keywording/stabling....in my eyes this is sufficient.
> Its obviously not, Many users are reporting file-collisions on a weekly
> basis. So either this isn't sufficient or the arch teams are not acting as
> you describe.

This can't be done yet, there's a couple of minor issues to iron out
(some python and gcc stuff collisions). After those few remaining
issues get finished, yeah this would be a useful default FEATURE.
There's quite a couple of bugs (usually pretty cryptic ones) caused by
clobbering files that belong to other ebuilds.

@Opfer: Users will usually notice many more of those collisions simply
because they are testing on a system that they normally use, while
developers are often testing in chroots or test boxes with just the
bare-bones system installed.

--

Jakub Moc
Email: jakub.moc@gmail.com
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: File collisions

[Alin Năstac]

[-- Attachment #1: Type: text/plain, Size: 946 bytes --]

Rob C wrote:
>
>
> On 19/04/07, *Christian Faulhammer* <opfer@gentoo.org
> <mailto:opfer@gentoo.org>> wrote:
>
>     Steve Long <slong@rathaus.eclipse.co.uk
>     <mailto:slong@rathaus.eclipse.co.uk>>:
>
>     > On the issue of QA, I think enabling FEATURES="collision-detect" by
>     > default would do a lot more good at this stage than "test".
>
>     Arch teams normally have collision-protect enabled when doing
>     keywording/stabling....in my eyes this is sufficient.
>
>     V-Li
>
>
> Its obviously not, Many users are reporting file-collisions on a
> weekly basis. So either this isn't sufficient or the arch teams are
> not acting as you describe.
>
The fact you can install a package using collision-detect is not a
guarantee that the package don't collide with some other package you
don't have installed on your test box (or with a package installed on
your box but not with the "right" USE flags).



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

[gentoo-dev] Re: File collisions

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 979 bytes --]

"Rob C" <hyakuhei@gentoo.org>:

> On 19/04/07, Christian Faulhammer <opfer@gentoo.org> wrote:
> > Steve Long <slong@rathaus.eclipse.co.uk>:
> > > On the issue of QA, I think enabling FEATURES="collision-detect"
> > > by default would do a lot more good at this stage than "test".
> > Arch teams normally have collision-protect enabled when doing
> > keywording/stabling....in my eyes this is sufficient.
> Its obviously not, Many users are reporting file-collisions on a
> weekly basis. So either this isn't sufficient or the arch teams are
> not acting as you describe.

 I can only speak for x86 team where it is mandatory for arch testers
and devs.  As far as I can tell other architecture teams handle it the
same.  Jakub already pointed out that many have very minimal chroots
for stabling, while my chroot is cluttered with 1500 packages which
should detect a lot of collisions.
 Just out of interest: How many file-collisions are filed per-week?

V-Li

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

[gentoo-dev] Re: File collisions

[Christian Faulhammer]

[-- Attachment #1: Type: text/plain, Size: 403 bytes --]

"Jakub Moc" <jakub.moc@gmail.com>:
> @Opfer: Users will usually notice many more of those collisions simply
> because they are testing on a system that they normally use, while
> developers are often testing in chroots or test boxes with just the
> bare-bones system installed.

 Yes.  Minimal chroots have the advantage to catch missing
dependencies.  So both have their pros and cons.

V-Li

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

Re: [gentoo-dev] Re: File collisions

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1010 bytes --]

On Fri, 2007-04-20 at 09:06 +0200, Rob C wrote:
> Arch teams normally have collision-protect enabled when doing
> keywording/stabling....in my eyes this is sufficient.
>         
>         V-Li
>         
> 
> Its obviously not, Many users are reporting file-collisions on a
> weekly basis. So either this isn't sufficient or the arch teams are
> not acting as you describe.

...or there's thousands of packages not going through the arch teams
because they haven't had new versions in $forever or the maintainer has
never asked for any version to go stable.  Again, I ask everyone to quit
being so confrontational and quit trying to prove other people wrong or
otherwise try to be argumentative just for the sake of being
argumentative.  We really don't need such behavior in what is supposed
to be a cooperative community.

-- 
Chris Gianelloni
Release Engineering Strategic Lead
Alpha/AMD64/x86 Architecture Teams
Games Developer/Council Member/Foundation Trustee
Gentoo Foundation

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

Re: [gentoo-dev] Re: File collisions

[Joshua Jackson]

[-- Attachment #1: Type: text/plain, Size: 1207 bytes --]

Rob C wrote:
>
>
> On 19/04/07, *Christian Faulhammer* <opfer@gentoo.org
> <mailto:opfer@gentoo.org>> wrote:
>
>     Steve Long <slong@rathaus.eclipse.co.uk
>     <mailto:slong@rathaus.eclipse.co.uk>>:
>
>     > On the issue of QA, I think enabling FEATURES="collision-detect" by
>     > default would do a lot more good at this stage than "test".
>
>     Arch teams normally have collision-protect enabled when doing
>     keywording/stabling....in my eyes this is sufficient.
>
>     V-Li
>
>
> Its obviously not, Many users are reporting file-collisions on a
> weekly basis. So either this isn't sufficient or the arch teams are
> not acting as you describe.
>
> -Rob
>
Rob,

Please watch it when saying that the arch teams are not acting as
described. I can tell you that we catch what comes to us. We don't get
every single package pushed on us as some never go through stable
testing, and we don't have every single package installed(that's
unrealistic). If they do then if there is a collision a note is filed in
the bug and we wait for a fix, as it actually does bail you out of the
build. This has been the mantra of at least x86 since the creation of
the team.


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

Re: [gentoo-dev] Re: File collisions

[Rob C]

[-- Attachment #1: Type: text/plain, Size: 1733 bytes --]

On 20/04/07, Joshua Jackson <tsunam@gentoo.org> wrote:
>
> Rob C wrote:
> >
> >
> > On 19/04/07, *Christian Faulhammer* <opfer@gentoo.org
> > <mailto:opfer@gentoo.org>> wrote:
> >
> >     Steve Long <slong@rathaus.eclipse.co.uk
> >     <mailto:slong@rathaus.eclipse.co.uk>>:
> >
> >     > On the issue of QA, I think enabling FEATURES="collision-detect"
> by
> >     > default would do a lot more good at this stage than "test".
> >
> >     Arch teams normally have collision-protect enabled when doing
> >     keywording/stabling....in my eyes this is sufficient.
> >
> >     V-Li
> >
> >
> > Its obviously not, Many users are reporting file-collisions on a
> > weekly basis. So either this isn't sufficient or the arch teams are
> > not acting as you describe.
> >
> > -Rob
> >
> Rob,
>
> Please watch it when saying that the arch teams are not acting as
> described. I can tell you that we catch what comes to us. We don't get
> every single package pushed on us as some never go through stable
> testing, and we don't have every single package installed(that's
> unrealistic). If they do then if there is a collision a note is filed in
> the bug and we wait for a fix, as it actually does bail you out of the
> build. This has been the mantra of at least x86 since the creation of
> the team.
>
>

It strikes me that people are often a little too sensative to any possible
doubts or aspersions that may be cast there way.

Taken in context I can't see why anyone would have a problem with what I
wrote. Either arch's are acting as described and the issue persists in which
case the action is not sufficient OR the pescribed action is sufficent but
not always undertaken or performed. I cant see how it can be both.

Regards
-Rob

[-- Attachment #2: Type: text/html, Size: 2536 bytes --]

Re: [gentoo-dev] Re: File collisions

[Steve Dibb]

> On 20/04/07, Joshua Jackson <tsunam@gentoo.org> wrote:
>>
>> Rob C wrote:
>> >
>> >
>> > On 19/04/07, *Christian Faulhammer* <opfer@gentoo.org
>> > <mailto:opfer@gentoo.org>> wrote:
>> >
>> >     Steve Long <slong@rathaus.eclipse.co.uk
>> >     <mailto:slong@rathaus.eclipse.co.uk>>:
>> >
>> >     > On the issue of QA, I think enabling FEATURES="collision-detect"
>> by
>> >     > default would do a lot more good at this stage than "test".
>> >
>> >     Arch teams normally have collision-protect enabled when doing
>> >     keywording/stabling....in my eyes this is sufficient.
>> >
>> >     V-Li
>> >
>> >
>> > Its obviously not, Many users are reporting file-collisions on a
>> > weekly basis. So either this isn't sufficient or the arch teams are
>> > not acting as you describe.
>> >
>> > -Rob
>> >
>> Rob,
>>
>> Please watch it when saying that the arch teams are not acting as
>> described. I can tell you that we catch what comes to us. We don't get
>> every single package pushed on us as some never go through stable
>> testing, and we don't have every single package installed(that's
>> unrealistic). If they do then if there is a collision a note is filed in
>> the bug and we wait for a fix, as it actually does bail you out of the
>> build. This has been the mantra of at least x86 since the creation of
>> the team.
>>
>>
>
> It strikes me that people are often a little too sensative to any possible
> doubts or aspersions that may be cast there way.
>
> Taken in context I can't see why anyone would have a problem with what I
> wrote. Either arch's are acting as described and the issue persists in
> which
> case the action is not sufficient OR the pescribed action is sufficent but
> not always undertaken or performed. I cant see how it can be both.

Well there is still the alternative option that the prescribed action is
sufficient, but it doesn't catch every corner case because that would be
unrealistic, so bugs are going to get reported.

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

Re: [gentoo-dev] Re: File collisions

[Matthias Langer]

On Fri, 2007-04-20 at 09:06 +0200, Rob C wrote:


> Its obviously not, Many users are reporting file-collisions on a
> weekly basis. So either this isn't sufficient or the arch teams are
> not acting as you describe. 

Can you provide some bug numbers to backup this claim?

Matthias

-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: File collisions

[Marijn Schouten (hkBst)]

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

Matthias Langer wrote:
> On Fri, 2007-04-20 at 09:06 +0200, Rob C wrote:
> 
> 
>> Its obviously not, Many users are reporting file-collisions on a
>> weekly basis. So either this isn't sufficient or the arch teams are
>> not acting as you describe. 
> 
> Can you provide some bug numbers to backup this claim?
> 
> Matthias

I count 33 open collision bugs
http://bugs.gentoo.org/buglist.cgi?quicksearch=collision

and 21 of those reported by users with a non-gentoo email.
http://tinyurl.com/3x9yt2

Marijn
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.3 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGKP7np/VmCx0OL2wRAvYhAJ4oK37gVQnDTIK8jHyijgh/aWMYEACbB3JM
G6BYMxkVRP3I8PCZFAC3isk=
=5DVn
-----END PGP SIGNATURE-----
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: File collisions

[Matthias Langer]

On Fri, 2007-04-20 at 19:56 +0200, Marijn Schouten (hkBst) wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Matthias Langer wrote:
> > On Fri, 2007-04-20 at 09:06 +0200, Rob C wrote:
> > 
> > 
> >> Its obviously not, Many users are reporting file-collisions on a
> >> weekly basis. So either this isn't sufficient or the arch teams are
> >> not acting as you describe. 
> > 
> > Can you provide some bug numbers to backup this claim?
> > 
> > Matthias
> 
> I count 33 open collision bugs
> http://bugs.gentoo.org/buglist.cgi?quicksearch=collision
> 
> and 21 of those reported by users with a non-gentoo email.
> http://tinyurl.com/3x9yt2
> 

Well, these are quite some bugs; however, at least the x86 arch team
(can't speak for the others, but i think they do it the same way) always
tests packages with "collision-protect". Since i'm an arch tester, i've
never seen that a package where we found collisions went to stable,
before these where fixed. Of course, we may have missed some collisions
every now and then, as it is in practice not possible to *ensure* that a
package has no collision with other packages.

As for enabling "collision-protect" by default: I'm not sure if this is
a good idea for now, as my experience is, that a significant part of the
packages that fail with "collision-protect" do so because of stale
files, that have been left around by (older versions of?) portage. As
soon as this is no longer the case, enabling "collision-protect" by
default sounds like a very good idea to me.

Matthias 

-- 
gentoo-dev@gentoo.org mailing list