James Manning

apt-file gets “command of the day”

% apt-file search /sbin/ip
iproute

today’s phrase: Post Hoc, Ergo Propter Hoc. That page gets the link since I like their Examples section

a little over 2 months to wait for it to hit cvs, but the simple end-of-line 3-line patch is now applied to log4j

hopefully security.debian.org will be back some time this weekend so I can get an SMP kernel going on this machine. Other system was p3/733, this one’s dual celery 366 @ 550’s

one thing i find funny is going to click logout at some site (and it’s been like an hour since i dealt with it) and it gives me a “session timed out, you must log in again to continue”. I know it’s somewhat silly to special-case that the logout path shouldn’t do the “timed out, relogin”, but still.

got the debian image over to the bp6 (sysresccd would have been nice if I had a bootable cd-rom drive, but debian’s rescue+root from images-1.44 worked fine too). Seems to be working fine so far.

I need to find the TNT2 Ultra that the p3/733 used to have, although maybe i can swap the scsi controller and drive (no longer needed) for a half-decent AGP card with someone.

so, a gentoo person came into #debian last night asking about building unstable from scratch (apt-build was the answer given for the curious). Others were annoying by a lot of *.debian.org machines not available but whatchagonnado.

so, from that and another conversation today:

base: genius
plural: genii (ok, not really)
roman: gen2
spelled out: gentwo
homophone: gentoo

don’t get me wrong, i’m sticking with Debian for the foreseeable future, but it was funny.

Course, it was fun to get into the “viruses, not virii” thread again. whee!

i’m gonna have to poke around to check (well, make sure), but if I open a file descriptor with O_SYNC (so explicit write calls are synchronous and blocking) and then mmap it, is that going to act any different than mmap’ing an fd that wasn’t opened with O_SYNC? My guess would be “no”, they both act the same and any desire to explicitly flush changes will still need an msync call. The alternative (all memory writes to that area are immediately flushed back to the file) is technically possible, but would be horrendous for performance :)

just kinda annoying that for “normal” writes, I can do sync writing with just O_SYNC and normal write (or pwrite or whatever) calls, but if I wanna take the mmap approach, I gotta include the explicit msync after a memcpy. I guess it’s just annoying to keep crossing the kernel/user boundary more often than I want. Ah, well.

I gotta get back to figuring out how I want that “do the mmap writes, but chunk it at 1GB each” loop to look like.

Speaking of which, props to whoever can first identify (w/o google or similar search methods) the sources (original ones, ideally) for the 2 test data 32-bit values in that test prog.

so i got the tester thingy working like I want it to, now to head back and get tiotest.c actually doing mmap correctly.