| 1 |
On 08/09/2012 01:19 PM, walt wrote: |
| 2 |
> This has been slow and painful so far. |
| 3 |
> |
| 4 |
> First, the build stops repeatedly because of zero-length library files. |
| 5 |
|
| 6 |
The gentoo libreoffice package doesn't yet compile with the 'binfilter' |
| 7 |
useflag set, so I unset it and the install completed. |
| 8 |
|
| 9 |
|
| 10 |
> Second, the ebuild is using only one CPU out of four, so this is taking |
| 11 |
> much longer than before. Ugh. |
| 12 |
|
| 13 |
This is a completely separate problem, which is very annoying but not |
| 14 |
fatal. There are some parts of the libreoffice build which take much |
| 15 |
longer than normal because 'make' uses 100% of CPU for long periods, |
| 16 |
while accomplishing very little. Those sections eventually complete, |
| 17 |
and the rest of the build proceeds at full speed by using all four of |
| 18 |
the CPU's. |
| 19 |
|
| 20 |
I think the reason that 'make' is being so slow is that certain kinds |
| 21 |
of child processes are segfaulting (i.e. become 'defunct') and make |
| 22 |
has to stop and respawn the failed processes until they succeed, which |
| 23 |
obviously takes a lot of extra time. The build does eventually finish |
| 24 |
in spite of all those defunct processes -- no idea why it works in the |
| 25 |
end, though. |
| 26 |
|
| 27 |
Can anyone confirm that 'make' is behaving the same way for you while |
| 28 |
building libreoffice-3.6.0.4? |
Archives