On 17/04/12 23:02, Robert Millan wrote: > El 14 de març de 2012 15:46, Christoph Egger <christ...@debian.org> ha > escrit: >> kern.dfldsiz: 134217728 > > That's just 128 MiB.
Argh I misunderstood this before. Yes 128 MiB is definitely too low for cmor; it seems to need ~700 MiB to pass the test suite or run at all. For kfreebsd-i386: cmor never built on this anyway, so it's not vital we fix that. Anyone planning to run climate models can surely find an amd64 system with plenty of RAM. It built on Linux i386, but only because it doesn't enforce this type of limit. The FreeBSD handbook mentions raising DFLDSIZ for "memory pigs"[1] but never higher than 1 GiB on i386. I think other kernel memory limits are hit at that point anyway. Trying to raise it much higher leaves my system unable to spawn init. With a DFLDSIZ setting of 512 MiB I can still boot a system with only 64 MiB physical RAM. To make the compiled-in default any higher would also require raising MAXDSIZ, or else it is still too low for cmor. So I suggest leaving this alone. For kfreebsd-amd64: A default of 128 MiB is quite low, and common software in Debian like mysql, squid or inkscape may have a problem with it. The MySQL 5.0 manual[2] recommends 1 GiB and I think that would be a sensible default. A DFLDSIZ of even 32 GiB on amd64 does not stop system with only 128 MiB physical RAM from booting. [1] http://www.freebsd.org/doc/handbook/sapr3.html#KERNELTUNING [2] https://dev.mysql.com/doc/refman/5.0/en/freebsd.html Regards, -- Steven Chamberlain ste...@pyro.eu.org -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
