| 1 |
On 06/17/2010 05:23 PM, Walter Dnes wrote: |
| 2 |
> On Thu, Jun 17, 2010 at 08:02:25AM -0700, Bill Longman wrote |
| 3 |
>> On 06/16/2010 05:33 PM, walt wrote: |
| 4 |
>>> On 06/16/2010 02:29 PM, Thomas Revell wrote: |
| 5 |
>>>> Hi everyone, |
| 6 |
>>>> |
| 7 |
>>>> I've got a bit of a problem with a new Gentoo install that I'm currently |
| 8 |
>>>> trying to install KDE on. The installation of kdebase/kfilereplace-4.3.5 |
| 9 |
>>>> is failing, apparently due to a missing header file in its sources. |
| 10 |
>>> |
| 11 |
>>> <snip> |
| 12 |
>>>> [ 11%] [34m[1mGenerating koptionsdlgs.h |
| 13 |
>>> <snip> |
| 14 |
>>>> [0mmoc: |
| 15 |
>>>> /var/tmp/portage/kde-base/kfilereplace-4.3.5/work/kfilereplace-4.3.5_build/kfilereplace/koptionsdlgs.h: |
| 16 |
>>>> |
| 17 |
>>>> No such file |
| 18 |
>>> |
| 19 |
>>> I have no idea what's going wrong, but I'm willing to make suggestions |
| 20 |
>>> anyway :) |
| 21 |
>> |
| 22 |
>> I have no better suggestion than to change your ricer CFLAGS and see if |
| 23 |
>> "-O2 -pipe -march=core2" works first. Especially since you have: |
| 24 |
>> |
| 25 |
>> "-mno-align-stringops -minline-stringops-dynamically" |
| 26 |
>> |
| 27 |
>> and kfilereplace.cpp:37 warns about QStringList& and further along, it |
| 28 |
>> creates kaddstringdlgs.h. |
| 29 |
>> |
| 30 |
>> BTW, -march=core2 implies -mmmx -msse -msse2 and -msse3 (and -mssse3). |
| 31 |
> |
| 32 |
> Even better is "-march=native", and let the compiler figure out what |
| 33 |
> features are available/safe. I use the following... |
| 34 |
> |
| 35 |
> CFLAGS="-O2 -march=native -mfpmath=sse -fomit-frame-pointer -pipe" |
| 36 |
> |
| 37 |
> If you're building 64-bit Gentoo, the "-mfpmath=sse" gets picked up by |
| 38 |
> "-march=native" and you can drop the explicit mention. |
| 39 |
> |
| 40 |
> And you wouldn't believe how many weird build problems are solved by... |
| 41 |
> |
| 42 |
> MAKEOPTS="-j1" |
| 43 |
> |
| 44 |
> ...even on multi-core cpus. The *BUILD PROCESS* is a bit slower, but |
| 45 |
> the final binary is identical with -j8 or whatever. And the time you |
| 46 |
> "save" by a faster build in a tty will be lost the first time you start |
| 47 |
> bashing your head into a brick wall over some weird build problem. |
| 48 |
> |
| 49 |
|
| 50 |
Exactly. And the same with distcc. If you do not use --keep-going, |
| 51 |
you'll likely see that long compile end abruptly. My typical modus |
| 52 |
operandii is to use distcc, then compile locally, then just use -j1 |
| 53 |
locally. You get there but you have to be willing to babysit it. |
| 54 |
|
| 55 |
For example, I just upgraded two of my VMs to gcc 4.4 so I will |
| 56 |
recompile "-qe world" on them. Since it's the weekend, I'll just turn |
| 57 |
off distcc and let them run -j1. When I come back on Monday, I expect to |
| 58 |
see a sparkling clean and shiny new Gentoo VM. Well, I can at least |
| 59 |
hope! I'll probably run out of inodes or something.... |
Archives