[gentoo-portage-dev] A way for trying to prevent useless rebuilds

[gentoo-portage-dev] A way for trying to prevent useless rebuilds

[Pacho Ramos]

This question comes to my mind every time a developer decides to
drop/add a USE flag to ebuilds like gcc/webkit-gtk/libreoffice... 

I think that we should have a file (like used for category movements) to
let PM know how to handle this situation. 

For example, 
category/foo-1.0 has a "gnome" USE flag but, later, that one is dropped:
-> If it is now *enabling* that support always, our file could have
something like:
category/foo gnome + -> that would mean that, when "gnome" USE flag is
NOT found, portage assumes it as being enabled, that will mean that
people having previously "gnome" enabled wouldn't need to rebuild the
package

... and the opposite

What do you think?

Thanks

Re: [gentoo-portage-dev] A way for trying to prevent useless rebuilds

[Zac Medico]

On 07/25/2013 11:54 AM, Pacho Ramos wrote:
> This question comes to my mind every time a developer decides to
> drop/add a USE flag to ebuilds like gcc/webkit-gtk/libreoffice... 
> 
> I think that we should have a file (like used for category movements) to
> let PM know how to handle this situation. 
> 
> For example, 
> category/foo-1.0 has a "gnome" USE flag but, later, that one is dropped:
> -> If it is now *enabling* that support always, our file could have
> something like:
> category/foo gnome + -> that would mean that, when "gnome" USE flag is
> NOT found, portage assumes it as being enabled, that will mean that
> people having previously "gnome" enabled wouldn't need to rebuild the
> package
> 
> ... and the opposite
> 
> What do you think?

We could do something like that. You should propose it in the gentoo-pms
list.
-- 
Thanks,
Zac

Re: [gentoo-portage-dev] A way for trying to prevent useless rebuilds

[Pacho Ramos]

El jue, 25-07-2013 a las 12:00 -0700, Zac Medico escribió:
> On 07/25/2013 11:54 AM, Pacho Ramos wrote:
> > This question comes to my mind every time a developer decides to
> > drop/add a USE flag to ebuilds like gcc/webkit-gtk/libreoffice... 
> > 
> > I think that we should have a file (like used for category movements) to
> > let PM know how to handle this situation. 
> > 
> > For example, 
> > category/foo-1.0 has a "gnome" USE flag but, later, that one is dropped:
> > -> If it is now *enabling* that support always, our file could have
> > something like:
> > category/foo gnome + -> that would mean that, when "gnome" USE flag is
> > NOT found, portage assumes it as being enabled, that will mean that
> > people having previously "gnome" enabled wouldn't need to rebuild the
> > package
> > 
> > ... and the opposite
> > 
> > What do you think?
> 
> We could do something like that. You should propose it in the gentoo-pms
> list.

OK, wanted to be sure a similar idea wasn't rejected before :)

Re: [gentoo-portage-dev] A way for trying to prevent useless rebuilds

[Zac Medico]

On 07/25/2013 12:14 PM, Pacho Ramos wrote:
> El jue, 25-07-2013 a las 12:00 -0700, Zac Medico escribió:
>> On 07/25/2013 11:54 AM, Pacho Ramos wrote:
>>> This question comes to my mind every time a developer decides to
>>> drop/add a USE flag to ebuilds like gcc/webkit-gtk/libreoffice... 
>>>
>>> I think that we should have a file (like used for category movements) to
>>> let PM know how to handle this situation. 
>>>
>>> For example, 
>>> category/foo-1.0 has a "gnome" USE flag but, later, that one is dropped:
>>> -> If it is now *enabling* that support always, our file could have
>>> something like:
>>> category/foo gnome + -> that would mean that, when "gnome" USE flag is
>>> NOT found, portage assumes it as being enabled, that will mean that
>>> people having previously "gnome" enabled wouldn't need to rebuild the
>>> package
>>>
>>> ... and the opposite
>>>
>>> What do you think?
>>
>> We could do something like that. You should propose it in the gentoo-pms
>> list.
> 
> OK, wanted to be sure a similar idea wasn't rejected before :)

I recall Brian Harring proposing something like that in the past, but it
never materialized.
-- 
Thanks,
Zac

Re: [gentoo-portage-dev] A way for trying to prevent useless rebuilds

[Alex Alexander]

[-- Attachment #1: Type: text/plain, Size: 1660 bytes --]

I once half-implemented something like this but never got to finish it,
I'll try to dig up my code and check out its state.

I recall using updates/ to tell portage what happened, which then updated
the VDB to reflect the new state.


On Thu, Jul 25, 2013 at 10:16 PM, Zac Medico <zmedico@gentoo.org> wrote:

> On 07/25/2013 12:14 PM, Pacho Ramos wrote:
> > El jue, 25-07-2013 a las 12:00 -0700, Zac Medico escribió:
> >> On 07/25/2013 11:54 AM, Pacho Ramos wrote:
> >>> This question comes to my mind every time a developer decides to
> >>> drop/add a USE flag to ebuilds like gcc/webkit-gtk/libreoffice...
> >>>
> >>> I think that we should have a file (like used for category movements)
> to
> >>> let PM know how to handle this situation.
> >>>
> >>> For example,
> >>> category/foo-1.0 has a "gnome" USE flag but, later, that one is
> dropped:
> >>> -> If it is now *enabling* that support always, our file could have
> >>> something like:
> >>> category/foo gnome + -> that would mean that, when "gnome" USE flag is
> >>> NOT found, portage assumes it as being enabled, that will mean that
> >>> people having previously "gnome" enabled wouldn't need to rebuild the
> >>> package
> >>>
> >>> ... and the opposite
> >>>
> >>> What do you think?
> >>
> >> We could do something like that. You should propose it in the gentoo-pms
> >> list.
> >
> > OK, wanted to be sure a similar idea wasn't rejected before :)
>
> I recall Brian Harring proposing something like that in the past, but it
> never materialized.
> --
> Thanks,
> Zac
>
>


-- 
Alex Alexander
+ wired
+ www.linuxized.com
+ www.leetworks.com

[-- Attachment #2: Type: text/html, Size: 2550 bytes --]

Re: [gentoo-portage-dev] A way for trying to prevent useless rebuilds

[Pacho Ramos]

El vie, 26-07-2013 a las 02:47 +0300, Alex Alexander escribió:
> I once half-implemented something like this but never got to finish
> it, I'll try to dig up my code and check out its state.
> 
> 
> I recall using updates/ to tell portage what happened, which then
> updated the VDB to reflect the new state.
> 

I have opened this same thread in pms-list, if you are able to forward
that code there, would be highly appreciated :)