SCM fetched recipes without a SRCREV - e.g. libgee-native

SCM fetched recipes without a SRCREV - e.g. libgee-native

[Koen Kooi]

Hi,

Lately some people have started adding recipes that fetch from an SCM 
without adding a srcrev either to the recipe or to sane-srcrevs.inc. 
That is bad. Unacceptable even.

e.g. 
http://dominion.thruhere.net/git/cgit.cgi/openembedded/commit/?id=cd473395ebf9adbf04ff4235878615cfa25daaf7

Previously it only affected recipes I don't care about, but the 
libgee-native one breaks my build. I seem to recall we declared taboo on 
recipes that have a floating srcrev by default, but I guess people think 
that "no srcrev at all" doesn't fit that category. It does.

So please add a srcrev or revert the cset.

Koen

Re: SCM fetched recipes without a SRCREV - e.g. libgee-native

[Holger Hans Peter Freyther]

On Wednesday 21 October 2009 12:47:41 Koen Kooi wrote:
> Hi,
> 
> Lately some people have started adding recipes that fetch from an SCM
> without adding a srcrev either to the recipe or to sane-srcrevs.inc.
> That is bad. Unacceptable even.

I agree, It must be possible to do

1.) bitbake -cfetchall task

then remove network and

2.) bitbake task...

if that is not the case we suck.

z.

Re: SCM fetched recipes without a SRCREV - e.g. libgee-native

[Michael 'Mickey' Lauer]

> 1.) bitbake -cfetchall task
> 
> then remove network and
> 
> 2.) bitbake task...
> 
> if that is not the case we suck.

Agreed. Perhaps it's really time to add some git commit checks as things
like the presence of a SRCREV_pn entry in sane-srcrev.inc can easily be
checked.

:M:

Re: SCM fetched recipes without a SRCREV - e.g. libgee-native

[Robert P. J. Day]

On Wed, 21 Oct 2009, Holger Hans Peter Freyther wrote:

> On Wednesday 21 October 2009 12:47:41 Koen Kooi wrote:
> > Hi,
> >
> > Lately some people have started adding recipes that fetch from an SCM
> > without adding a srcrev either to the recipe or to sane-srcrevs.inc.
> > That is bad. Unacceptable even.
>
> I agree, It must be possible to do
>
> 1.) bitbake -cfetchall task
>
> then remove network and
>
> 2.) bitbake task...
>
> if that is not the case we suck.

  *yes*, that's what i'd forgotten.  i'd tried just

  $ bitbake -c fetch ...

but after i started a bitbake, there was still some downloading.  so
how much extra work does "fetchall" do as a command as opposed to
just "fetch"?  thanks.

rday
--

========================================================================
Robert P. J. Day                               Waterloo, Ontario, CANADA

        Linux Consulting, Training and Annoying Kernel Pedantry.

Web page:                                          http://crashcourse.ca
Twitter:                                       http://twitter.com/rpjday
========================================================================

Re: SCM fetched recipes without a SRCREV - e.g. libgee-native

[Koen Kooi]

On 21-10-09 13:10, Michael 'Mickey' Lauer wrote:
>> 1.) bitbake -cfetchall task
>>
>> then remove network and
>>
>> 2.) bitbake task...
>>
>> if that is not the case we suck.
>
> Agreed. Perhaps it's really time to add some git commit checks as things
> like the presence of a SRCREV_pn entry in sane-srcrev.inc can easily be
> checked.

If we continue down the sane-srcrevs.inc route we should consider moving 
PV_pn into there as well. Having SRCREV and PN in 2 different places 
only adds confusion and leads to misversions.
Speaking of misversioning, I should really go through all scm fetched 
packages I use and verify PV is correct for those.

regards,

Koen

Re: SCM fetched recipes without a SRCREV - e.g. libgee-native

[Michael 'Mickey' Lauer]

Am Mittwoch, den 21.10.2009, 13:21 +0200 schrieb Koen Kooi:
> On 21-10-09 13:10, Michael 'Mickey' Lauer wrote:
> >> 1.) bitbake -cfetchall task
> >>
> >> then remove network and
> >>
> >> 2.) bitbake task...
> >>
> >> if that is not the case we suck.
> >
> > Agreed. Perhaps it's really time to add some git commit checks as things
> > like the presence of a SRCREV_pn entry in sane-srcrev.inc can easily be
> > checked.
> 
> If we continue down the sane-srcrevs.inc route we should consider moving 
> PV_pn into there as well. Having SRCREV and PN in 2 different places 
> only adds confusion and leads to misversions.

Correct.

> Speaking of misversioning, I should really go through all scm fetched 
> packages I use and verify PV is correct for those.

Agreed, will have to do the same here.

Regards,

:M:

Re: SCM fetched recipes without a SRCREV - e.g. libgee-native

[Richard Purdie]

On Wed, 2009-10-21 at 13:42 +0200, Michael 'Mickey' Lauer wrote:
> Am Mittwoch, den 21.10.2009, 13:21 +0200 schrieb Koen Kooi:
> > If we continue down the sane-srcrevs.inc route we should consider moving 
> > PV_pn into there as well. Having SRCREV and PN in 2 different places 
> > only adds confusion and leads to misversions.
> 
> Correct.
> 
> > Speaking of misversioning, I should really go through all scm fetched 
> > packages I use and verify PV is correct for those.
> 
> Agreed, will have to do the same here.

Can I also suggest we start using SRCPV in PV, not SRCREV? This removed
the need to play a lot of games with PV...

(Poky has switched to this now)

Cheers,

Richard

Re: SCM fetched recipes without a SRCREV - e.g. libgee-native

[Denys Dmytriyenko]

On Wed, Oct 21, 2009 at 07:57:40PM +0100, Richard Purdie wrote:
> On Wed, 2009-10-21 at 13:42 +0200, Michael 'Mickey' Lauer wrote:
> > Am Mittwoch, den 21.10.2009, 13:21 +0200 schrieb Koen Kooi:
> > > If we continue down the sane-srcrevs.inc route we should consider moving 
> > > PV_pn into there as well. Having SRCREV and PN in 2 different places 
> > > only adds confusion and leads to misversions.
> > 
> > Correct.
> > 
> > > Speaking of misversioning, I should really go through all scm fetched 
> > > packages I use and verify PV is correct for those.
> > 
> > Agreed, will have to do the same here.
> 
> Can I also suggest we start using SRCPV in PV, not SRCREV? This removed
> the need to play a lot of games with PV...
> 
> (Poky has switched to this now)

AFAIK, it's not merged in OE yet. I would definitely like to see it in OE.
Last time we discussed it, the agreement was to introduce SRCPV in 
bitbake.conf first and then gradually switch existing recipes over from 
SRCREV, instead of mass-updating all of them at once...

-- 
Denys

Re: SCM fetched recipes without a SRCREV - e.g. libgee-native

[Richard Purdie]

On Wed, 2009-10-21 at 20:59 -0400, Denys Dmytriyenko wrote:
> On Wed, Oct 21, 2009 at 07:57:40PM +0100, Richard Purdie wrote:
> > On Wed, 2009-10-21 at 13:42 +0200, Michael 'Mickey' Lauer wrote:
> > > Am Mittwoch, den 21.10.2009, 13:21 +0200 schrieb Koen Kooi:
> > > > If we continue down the sane-srcrevs.inc route we should consider moving 
> > > > PV_pn into there as well. Having SRCREV and PN in 2 different places 
> > > > only adds confusion and leads to misversions.
> > > 
> > > Correct.
> > > 
> > > > Speaking of misversioning, I should really go through all scm fetched 
> > > > packages I use and verify PV is correct for those.
> > > 
> > > Agreed, will have to do the same here.
> > 
> > Can I also suggest we start using SRCPV in PV, not SRCREV? This removed
> > the need to play a lot of games with PV...
> > 
> > (Poky has switched to this now)
> 
> AFAIK, it's not merged in OE yet. I would definitely like to see it in OE.
> Last time we discussed it, the agreement was to introduce SRCPV in 
> bitbake.conf first and then gradually switch existing recipes over from 
> SRCREV, instead of mass-updating all of them at once...

Either way is fine but this looks like a good time to think about it if
we're already assessing the variables!

Cheers,

Richard