dry-run defunct

dry-run defunct

[Pjotr Prins]

Did --dry-run stop working, or is it me? Basically the flag is ignored
on the latest checkout.

Pj.

Re: dry-run defunct

[Andreas Enge]

On Sun, Jul 17, 2016 at 12:32:18PM +0200, Pjotr Prins wrote:
> Did --dry-run stop working, or is it me? Basically the flag is ignored
> on the latest checkout.

It does not quite work in the presence of grafts. As far as I understand
from testing, the ungrafted packages are still built, and then it stops
only before grafting.

Andreas

Re: dry-run defunct

[Ludovic Courtès]

Andreas Enge <andreas@enge.fr> skribis:

> On Sun, Jul 17, 2016 at 12:32:18PM +0200, Pjotr Prins wrote:
>> Did --dry-run stop working, or is it me? Basically the flag is ignored
>> on the latest checkout.
>
> It does not quite work in the presence of grafts. As far as I understand
> from testing, the ungrafted packages are still built, and then it stops
> only before grafting.

Specifically, to determine whether a binary needs to be grafted, Guix
must check whether that binary depends on one of the things that we
would graft.

To do that, it must either have access to substitute information, or
have a local build of the thing at hand.

Thus, when substitutes are lacking, it will tend to build things right
away so that it can answer that question.

I agree this isn’t great, but I’m not sure how to improve on that.
(Nevertheless, it is better than answering the question based on
compile-time dependencies as was done before.)

Ludo’.

Re: dry-run defunct

[Pjotr Prins]

On Sun, Jul 17, 2016 at 03:40:56PM +0200, Ludovic Courtès wrote:
> I agree this isn’t great, but I’m not sure how to improve on that.
> (Nevertheless, it is better than answering the question based on
> compile-time dependencies as was done before.)

Grafts, I thought, an intermediate solution to prevent excessive
rebuilds. Still, regularly we would rebuild everything, right? Or are
we now grafting on grafting?

Pj.
-- 

Re: dry-run defunct

[Andreas Enge]

On Sun, Jul 17, 2016 at 06:52:01PM +0200, Pjotr Prins wrote:
> Grafts, I thought, an intermediate solution to prevent excessive
> rebuilds. Still, regularly we would rebuild everything, right? Or are
> we now grafting on grafting?

Yes. No. :-)  We undo the grafts in core-updates and rebuild everything
there.

Andreas

Re: dry-run defunct

[Ludovic Courtès]

Andreas Enge <andreas@enge.fr> skribis:

> On Sun, Jul 17, 2016 at 06:52:01PM +0200, Pjotr Prins wrote:
>> Grafts, I thought, an intermediate solution to prevent excessive
>> rebuilds. Still, regularly we would rebuild everything, right? Or are
>> we now grafting on grafting?
>
> Yes. No. :-)  We undo the grafts in core-updates and rebuild everything
> there.

Exactly.  So far we did not rebuild as frequently as we’d like due to
limited resources, but hopefully the situation will improve when our new
server is up and running!

Ludo’.