| 1 |
Am Mittwoch, 4. Februar 2009 21:21:38 schrieb Alan McKinnon: |
| 2 |
> On Wednesday 04 February 2009 20:17:33 Dirk Heinrichs wrote: |
| 3 |
> > Am Mittwoch, 4. Februar 2009 04:25:34 schrieb ABCD: |
| 4 |
> > > The reason there wasn't a bump (IIRC) was that the ebuild never changed |
| 5 |
> > > - only the eclass did. If you emerged any version of GCC during the |
| 6 |
> > > window where the eclass was broken, that version of GCC would have been |
| 7 |
> > > broken. |
| 8 |
> > |
| 9 |
> > That also means that portage is broken. Whenever something changes where |
| 10 |
> > other things depend on, those other things should be rebuilt. This |
| 11 |
> > doesn't happen here. |
| 12 |
> |
| 13 |
> I disagree, that would cause many more spurious rebuilds than is needed if |
| 14 |
> it were automated. |
| 15 |
|
| 16 |
Why spurious? The package manager should be smart enough to tell the user: |
| 17 |
"Rebuild because of eclass change". Nothing spurious. |
| 18 |
|
| 19 |
> Portage cannot tell how important or how deep a change |
| 20 |
> is, that requires a human to look and see. So what is needed is a flag that |
| 21 |
> portage recognizes as an instruction to do a revdep-rebuild-style search |
| 22 |
> and find everything using a specific changed file, and rebuild those. The |
| 23 |
> flag is set under dev control. |
| 24 |
|
| 25 |
Not dev, user. Something equivalent to --newuse. |
| 26 |
|
| 27 |
> Blindly doing what you suggest leads to this: |
| 28 |
> |
| 29 |
> appA depends on libB. |
| 30 |
|
| 31 |
No. Sorry if this was misleading. I was only referring to dependencies between |
| 32 |
ebuilds and eclasses. |
| 33 |
|
| 34 |
Library interdependencies are handled just fine by portage/revdep-rebuild. |
| 35 |
|
| 36 |
> Therefore, a simple elog entry is a valid handling and fully compliant with |
| 37 |
> the principle of The Simplest Thing That Could Possibly Work. |
| 38 |
|
| 39 |
Elog entries are overlooked too often. |
| 40 |
|
| 41 |
Bye... |
| 42 |
|
| 43 |
Dirk |
Archives