When glibc is built on a distcc system, the compile is corectly distributed. When the install comes to building locale data, it is ALSO built with whatever makeopts are set (MAKEOPTS="-j4" for me) this is *very* *very* slow since it is not distributed via distcc. Is it possible to check for distcc in the ebuild and not use MAKEOPTS when building locale data?
So having it run as -j1 on the same machine, is quicker than -j4 if distcc is running ?
Yes, but *only* for the locale building.
please cc me on distcc issues. Personally this sounds like a Portage or Makefile thing.
Yes it is, portage *thinks* that MAKEOPTS="-j4" means it can build 4 locales at a time. Locale building is running a program, not using gcc, so the intended distcc distribution does not happen, and the system is heavily taxed and builds locales very slowly.
In my case, "very slowly" means 12 locales in 48 HOURS. Four hours per locale is rediculous, even for a 333Mhz PII. Currently the only way I've found around this during installation is to not distribute during bootstrapping, and then MAKEOPTS="-j1" emerge glibc before emerging system.
Seems to be no movement here... Isn't it just a matter of making the glibc ebuild use a single locale build at a time.
I wanted to emerge the latest glibc using distcc and it gave an error during the process, when i disabled distcc it emerged fine. Maybe interesting to do a distcc check before emerging and if needed not using distcc Filip
lv: maybe it's just me, but installing in parallel is a bad idea perhaps install_locales() should be updated to read: PARALLELMFLAGS="${MAKEOPTS} -j1" instead of: PARALLELMFLAGS="${MAKEOPTS}" can the people with this problem verify this fix ?
I had that problem a while ago and added in that fix, so closing.
