| 1 |
On Thursday 13 May 2010 06:06:17 Walter Dnes wrote: |
| 2 |
> On Mon, May 10, 2010 at 12:33:21AM +0100, Neil Bothwick wrote |
| 3 |
> |
| 4 |
> > On Sun, 9 May 2010 18:28:31 -0400, Walter Dnes wrote: |
| 5 |
> > > Ditto for setting MAKEOPTS to -j1. Every once in a while, somebody |
| 6 |
> > > |
| 7 |
> > > runs into a problem that is solved by it. I finally decided to let |
| 8 |
> > > the builds take a little bit longer, in exchange for saving me problems |
| 9 |
> > > with "unreproducable errors". This setting does not affect the final |
| 10 |
> > > binary; just how long it takes to build. |
| 11 |
> > |
| 12 |
> > I can't recall the last time I needed to use MAKEOPTS=-j1, but if you |
| 13 |
> > do set it you can get back the time you lose by using the jobs option |
| 14 |
> > with emerge. That way you get parallel compilation, but of separate |
| 15 |
> > packages. |
| 16 |
> |
| 17 |
> Something just occured to me. At the risk of sounding paranoid, is |
| 18 |
> there an absolute guarantee that... |
| 19 |
> - if package A has dependancies Y and Z... |
| 20 |
> - the compile for A won't start before Y and Z are built and installed |
| 21 |
|
| 22 |
Yes. This is easy to do, IIRC Python even ships with the necessary library |
| 23 |
routines for the data structures required. |
| 24 |
|
| 25 |
One builds a data graph that represents dependencies, when finished you are |
| 26 |
guaranteed that if the graph is read from the bottom up then your conditions |
| 27 |
are met. |
| 28 |
|
| 29 |
Elementary my dear Watson, we cover this in first year Comp Sci courses |
| 30 |
|
| 31 |
> This could be especially ugly for my new system installs. I usually |
| 32 |
> install text-console mode only, followed by "emerge gimp", which pulls |
| 33 |
> in X and a whole bunch of other stuff as dependancies. Portage handles |
| 34 |
> this process very well right now. |
| 35 |
|
| 36 |
It is this way because Gentoo is unusuable without it |
| 37 |
|
| 38 |
|
| 39 |
-- |
| 40 |
alan dot mckinnon at gmail dot com |
Archives