| 1 |
Finn Thain wrote: |
| 2 |
>> When everything goes fine, no email notification is being sent out. A |
| 3 |
>> convenient log structure would, however, make it possible to see which |
| 4 |
>> packages and USE-flag combinations successfully passed through. |
| 5 |
>> Providing this log via a web-page would be a useful thing. |
| 6 |
> |
| 7 |
> Would tinderbox help? |
| 8 |
|
| 9 |
As far as I know about tinderbox, it is more than just a building |
| 10 |
system. It is a complete procedure where the tree is being closed |
| 11 |
during compilation time, then only reopened when everything compiles. |
| 12 |
|
| 13 |
> |
| 14 |
>> - Comments are welcome, as well as expressions of worry on my mental state. |
| 15 |
> |
| 16 |
> Good thinking! |
| 17 |
> |
| 18 |
> The chroot idea is a good one because the process lends itself to |
| 19 |
> parallelism. That is, you might have one test box/chroot for, (maybe in |
| 20 |
> order of importance) |
| 21 |
> |
| 22 |
> - unstable empty tree (all deps every time) |
| 23 |
> - stable empty tree builds (same) |
| 24 |
> - unstable cumulative tree builds |
| 25 |
> - stable cumulative tree builds |
| 26 |
|
| 27 |
This is indeed a good plan, as this allows some more responsive and |
| 28 |
thorough testing to occur next to each other. |
| 29 |
|
| 30 |
> I see the last ones as being fairly important, because the cumulative |
| 31 |
> (emerge -Du) trees will have the best throughput, for quicky finding any |
| 32 |
> glaring, slap-forehead kind of bugs/bad keywords (i.e. low fruit). |
| 33 |
> |
| 34 |
> The cumulative tree machines would also be an efficient choice for your |
| 35 |
> reverse-dependency idea (perhaps to only one level of indirection). |
| 36 |
|
| 37 |
Good points, thanks! |
| 38 |
|
| 39 |
-- |
| 40 |
Fabian Groffen |
| 41 |
eBuild && Porting |
| 42 |
Gentoo for Mac OS X |
| 43 |
-- |
| 44 |
gentoo-osx@g.o mailing list |
Archives