| 1 |
On Sun, 2004-11-07 at 20:58 +0000, Ciaran McCreesh wrote: |
| 2 |
> On Sun, 07 Nov 2004 21:54:07 +0100 Patrick Lauer <patrick@g.o> |
| 3 |
> wrote: |
| 4 |
> | after the recent discussion about eclass versioning, possible flaws in |
| 5 |
> | eclass signing/authentication I'd like to see eclass versioning |
| 6 |
> | implemented. This will hopefully reduce some frustrating bugs and |
| 7 |
> | consistency problems, but will cause a bit more work for the |
| 8 |
> | maintainers and some small additions to portage. |
| 9 |
> |
| 10 |
> Well, you can handle incompatible package versions from the one eclass |
| 11 |
> using versionator. toolchain and vim do this already... |
| 12 |
|
| 13 |
Yes, no problems with that, but I was thinking about incompatible |
| 14 |
_eclass_ versions. At the moment, no easy checks for changed eclasses |
| 15 |
are available. This has even be compared to the windows "dll hell" by |
| 16 |
some people since it's possible but unneccessarily difficult to keep |
| 17 |
your system consistent when eclasses (rarely as that happens) change. |
| 18 |
(Or rather, their behaviour changes in (for the user) unpredictable |
| 19 |
ways). Also, what happens if someone uses an overlay with a |
| 20 |
better/newer/older eclass version? As long as there is no distinction |
| 21 |
between versions, I can imagine lots of *ahem* interesting problems that |
| 22 |
could be avoided. |
| 23 |
|
| 24 |
Thanks for your input, |
| 25 |
|
| 26 |
Patrick |
Archives