| 1 |
On Thursday 27 September 2007, Christian Faulhammer wrote: |
| 2 |
> Mike Frysinger <vapier@g.o>: |
| 3 |
> > On Thursday 27 September 2007, Christian Faulhammer wrote: |
| 4 |
> > > could there be side effects setting MAKEOPTS in the ebuild (in |
| 5 |
> > > global scope)? |
| 6 |
> > |
| 7 |
> > you should only be incrementing, never setting ... anything that sets |
| 8 |
> > it is broken while anything that increments it is considered bad |
| 9 |
> > form, but currently overlooked due to a better solution lacking |
| 10 |
> > iow, wrong: |
| 11 |
> > MAKEOPTS="-j1" |
| 12 |
> > currently overlooked: |
| 13 |
> > MAKEOPTS="${MAKEOPTS} -j1" |
| 14 |
> |
| 15 |
> Ok, so the latter saves me from repeating all standard functions with |
| 16 |
> emake -j1. Thanks. |
| 17 |
|
| 18 |
sorry, i'll be a bit more specific ... the reason appending MAKEOPTS has been |
| 19 |
sliding is due to the large volume of gnome ebuilds and integrated |
| 20 |
eclasses ... it isnt trivial to pass that information along since the emake |
| 21 |
happens in nested eclasses, so MAKEOPTS gets incremented in the ebuild |
| 22 |
|
| 23 |
if you're doing it to avoid -j1 in like two calls to emake, that's poor |
| 24 |
taste :p |
| 25 |
|
| 26 |
also, any package that fails -j1 building should have an open bug about it ... |
| 27 |
parallel building isnt something to be workedaround with -j1 and then ignored |
| 28 |
as it can severely screw people (consider systems that have a crap ton of |
| 29 |
slow procs -- common for mips people) |
| 30 |
-mike |
Archives