Re: lang/rust PORT_LLVM=on Dependency loop detected

On Fri, 26 Jan 2024 23:13:34 -0500
Charlie Li <vishwin@freebsd.org> wrote:

> Miroslav Lachman wrote:
> > I tried to build our package set (in Poudriere) with the new option 
> > PORT_LLVM=on for lang/rust, but it seems it does not work / has not been 
> > tested, or we have some other non-default options that cause the 
> > following error:
> > 
> > [00:00:57] Error: Dependency loop detected:
> > These packages depend on each other: py39-openssl-23.2.0,1 
> > py39-cryptography-41.0.7_1,1 rust-1.74.1 llvm15-15.0.7_9 
> > py39-recommonmark-0.5.0_2 py39-sphinx-5.3.0_1,1 py39-requests-2.31.0 
> > py39-urllib3-1.26.18,1
> > These packages depend on each other: rust-1.74.1 llvm15-15.0.7_9 
> > py39-recommonmark-0.5.0_2 py39-sphinx-5.3.0_1,1 py39-requests-2.31.0 
> > py39-urllib3-1.26.18,1 py39-openssl-23.2.0,1 
> > py39-cryptography-41.0.7_1,1 py39-setuptools-rust-1.8.1
> > 
> You need to disable the SSL option in net/py-urllib3, see 
> https://cgit.freebsd.org/ports/commit/?id=761dc8a9f4ddef2adad122a57a477e65457582fb
> 
> -- 
> Charlie Li
> ...nope, still don't have an exit line.

Hi.
Maybe this would be caused by "changes on default option" problem on
DIALOG4PORTS (incuding both dialog4ports and portconfig) behaviour.

When NEW options are ADDED, `make config-recursive` shows dialog for
it, marking them as "New". But when any of options are removed, nothing
happens. And on changes in defaults case, dialog appears but no advices
are shown.

This makes it difficult to track changes. When dialog appears but no
"New" marker is shown, I habitally look into commit diff with posts on
dev-commits-ports-main ML or cgit to see what changed.
But although doing so, I miss changes of defaults if addition of new
option and changes to default options are done in single commit.

It would be very nice if...

  *Options that changed default are somehow marked,

or

 *Somehow mark every option which has non-default value.

Not sure how difficult it is, but either of them would help for this
kind of regressions.

Regards.

-- 
Tomoaki AOKI    <junchoon@dec.sakura.ne.jp>