1 |
On Wed, 14 Aug 2019 11:42:16 +1200 |
2 |
Kent Fredric <kentnl@g.o> wrote: |
3 |
|
4 |
> On Tue, 13 Aug 2019 20:39:01 +0100 |
5 |
> James Le Cuirot <chewi@g.o> wrote: |
6 |
> |
7 |
> > Unfortunately there's currently no way to say that the versions of a |
8 |
> > package across BDEPEND, DEPEND, and RDEPEND must be (roughly?) equal so |
9 |
> > there's nothing to stop an ancient Perl in / building a module for a |
10 |
> > newer Perl in ROOT but that's a problem that goes far beyond Perl. |
11 |
> > Cross-compiling Perl modules is still a bit of a car crash anyway. |
12 |
> |
13 |
> I'd imagine that for CC, you'd need all perl modules installed to |
14 |
> BDEPEND _and_ DEPEND _AND_ RDEPEND, |
15 |
> |
16 |
> Because the configure stage runs Makefile.PL on the host perl, and it |
17 |
> checks the existence of all "runtime" and "compile time" dependencies |
18 |
> itself, and it does this by *loading* them into the host perl. |
19 |
> |
20 |
> Failure to load them on the host perl produces lots of warnings in the |
21 |
> best of cases, and errors in the worst of cases. |
22 |
|
23 |
Yes, I thought that was probably the case. Python is more forgiving in |
24 |
this regard. |
25 |
|
26 |
When trying to solve this problem in cross-boss, I found that |
27 |
Makefile-based modules would usually build if you ran Perl through QEMU |
28 |
during src_configure and ran Make with some tweaked variables during |
29 |
src_compile and src_install but this is extremely hacky. |
30 |
|
31 |
-- |
32 |
James Le Cuirot (chewi) |
33 |
Gentoo Linux Developer |