Bug#1027686: transition: rakudo
Control: tags -1 moreinfo
On 2023-01-09 13:54:08 -0500, M. Zhou wrote:
> I missed the detail that the compiler ID even changes for different
> architecture.. which may not be good.
Is it required that the build path of the compiler influences the ID?
Can the computation of the ID be patched to be independent of the
build path?
> Is it possible for us to slightly modify the postinst script to
> recompile the cache locally when the compiler id mismatches?
> The fallback script rakudo-helper.pl can at least make sure
> a raku-* package is still functional even without a matching
> compiler ID. In that case we don't have to add the compiler ID
> to the virtual package name, and every architecture can track
> the same and consistent virtual package dependency.
I think it's too late to change this shortly before the freeze starts.
If fixing the compiler ID computation is not possible, I prefer to delay
this transition to trixie.
Cheers
> On Sat, 2023-01-07 at 18:40 +0100, Dominique Dumont wrote:
> > On Saturday, 7 January 2023 11:58:29 CET you wrote:
> > > > Unfortunately, the compiler-id also depends on the build
> > > > directory. Which
> > > > means that the compiler id changes between arches.
> > >
> > > This should be fixed first. Otherwise every rebuild of the compiler
> > > will
> > > require all reverse dependencies to be rebuilt too. That does not
> > > sound
> > > like a good solution.
> >
> > Agreed, but that's a long story with upstream:
> >
> > https://github.com/rakudo/rakudo/issues/5099
> >
> > All the best
> >
>
--
Sebastian Ramacher
Reply to: