| 1 |
On Sun, 2019-05-05 at 17:10 +0200, Ulrich Mueller wrote: |
| 2 |
> > > > > > On Sun, 05 May 2019, Michał Górny wrote: |
| 3 |
> > Subject: [PATCH] ebuild-env-vars.text: Allow SYSROOT & BROOT in pkg_setup |
| 4 |
> |
| 5 |
> s/\.text/.tex/ |
| 6 |
> |
| 7 |
> > Allow using SYSROOT & BROOT in pkg_setup(). This follows the earlier |
| 8 |
> > change allowing for build-time dependencies in pkg_setup, whenever |
| 9 |
> > it is being run as part of source build. |
| 10 |
> |
| 11 |
> That doesn't look right, conceptually. It is not guaranteed that any |
| 12 |
> build time dependencies will be still present in the BROOT seen by |
| 13 |
> pkg_setup, when merging a binary package. |
| 14 |
> |
| 15 |
> We could add a footnote saying that it's only available when pkg_setup |
| 16 |
> is executed as part of a source build, but that looks way to complicated |
| 17 |
> and error prone. |
| 18 |
> |
| 19 |
|
| 20 |
The alternative is to either ignore the fact that eclasses will be doing |
| 21 |
it, or force eclasses to drop established API for the sake of |
| 22 |
compatibility with misdesigned specification. And yes, this all boils |
| 23 |
down to missing src_setup(). |
| 24 |
|
| 25 |
-- |
| 26 |
Best regards, |
| 27 |
Michał Górny |
Archives