| 1 |
> j2 for MAKEOPTS and +kdeenablefinal and after some time each one of the |
| 2 |
> makejobs want 900mb ram. There are two libs where that happens, makes |
| 3 |
> kdepim the slowest-to-compile packet for me. Wesnoth is also an offender. |
| 4 |
> Some versions want 500mb+ at some point when compiling. |
| 5 |
|
| 6 |
That's pretty much. About 3 years ago I watched a similar behaviour on PyQt. |
| 7 |
At that time I had just 512MB of ram (and a 1.2GHz pentium3) and that tiny |
| 8 |
package drove me crazy because the memory requested by gcc when compiling |
| 9 |
PyQt was more than 512MB and the compile process did not only happen on data |
| 10 |
stored in the ram but also on data stored in the swap space. It was awful. My |
| 11 |
system was not usable for more than an hour. It was so busy swapping as I |
| 12 |
never saw a system swapping before. I remember that after a upgrade to 1G ram |
| 13 |
the PyQt package took less than five minutes. I decided to trace the memory |
| 14 |
usage when compiling PyQt and I recognized that it was less than a minute |
| 15 |
requesting so much memory. That was the first time I was convinced that |
| 16 |
swapping can end very very bad. |
| 17 |
|
| 18 |
rgds |
| 19 |
Bernhard |
| 20 |
|
| 21 |
-- |
| 22 |
gentoo-amd64@g.o mailing list |
Archives