Created attachment 308805 [details] cdk-5.0.20090215-build.log I do not encouner the errors about a missing fscale.h in the attached build.log when I do: MAKEOPTS= emerge -1v cdk Also, I have only been able to reproduce the parallel compilation failure when taxing the system by merging something else (ex. freeciv) at the same time.
Created attachment 308807 [details] emerge--info.txt
Can you reproduce this with 5.0.20110517 or 5.0.20120323?
Created attachment 308853 [details] cdk-5.0.20120323-build.log (In reply to comment #2) > Can you reproduce this with 5.0.20110517 or 5.0.20120323? Yup.
Created attachment 308859 [details, diff] parallel make patch Since I can't reproduce the failure you'll have to test the patch to see if it fixes your problem. Just append it to the other two patches already applied in cdk-5.0.20120323.
Created attachment 308861 [details, diff] parallel make patch Attached the correct patch this time.
Created attachment 308911 [details] cdk-5.0.20120323-build.log attachment 308861 [details, diff] does not affect the build. Your patch makes certain invalid assumptions about what is needed to fix parallel build problems. all: $(AUTO_HDR) $(DEFAULT) The above line does not guarantee that $(AUTO_HDR) will be built before $(DEFAULT) is. You should instead need to have any object file whose source file requires the AUTO_HDRs to depend on AUTO_HDRs. (patch coming in a moment)
Created attachment 308913 [details, diff] cdk-5.0.20120323-parallel-make.patch This version makes the object files depend on _all_ of the headers instead of just the one non-generated header. It seems to fix the problem for me. And its logic seems right ;-). You said you had trouble reproducing the bug. I also do, but I found that I was able to relatively reliably reproduce it by making sure that freeciv was in src_compile() before starting to emerge cdk. I'm not going to try to guess why that works for me, but that's how parallel stuff goes probably…
> You said you had trouble reproducing the bug. I also do, but I found that I > was able to relatively reliably reproduce it by making sure that freeciv was > in src_compile() before starting to emerge cdk. I'm not going to try to > guess why that works for me, but that's how parallel stuff goes probably… I tried a lot of different cases and none of them failed for me.
(In reply to comment #7) > Created attachment 308913 [details, diff] [details, diff] > cdk-5.0.20120323-parallel-make.patch I'm assuming by the looks of your patch you sent this upstream? If not, please do.
Fixed in CVS.
