Camelia, the Perl 6 bug

IRC log for #etools, 2012-04-18

| Channels | #etools index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:36 richardus joined #etools
01:38 richardus hi guys, i'm working on a ghetto reimplemetation of entwatch/entmesh in python
01:38 richardus i'll give this channel a heads up when i have things on github!
02:01 mikemol richardus: Cool
02:02 mikemol FWIW, entmesh hasn't been implemented yet.
02:02 mikemol So no wire protocol has really been defined, just signalling semantics.
02:05 mikemol richardus: Also, you probably ought to read some stuff I haven't implemented yet. http://irclog.perlgeek.de/e​tools/2012-04-17#i_5455984
02:07 richardus interesting.  btw when did you roll out the update for ioctl vs fwrite - i iamgine you need to throw out the old data?
02:07 richardus my (prgmr) vps doesn't seem to get that much traffic and thus doesn't have as much entropy.  if i notice something notable i'll toss it out here for the perspective
02:08 richardus yeah, i took a stab at it over the weekend.  i'm new to network programming in general so i understand i'll probably have to rewrite stuff when you have an implementation.  that's also why i opened the bug on the text description - i was having a hard time coming up with exactly what the document was telling me i had to support
02:09 richardus oh whoops - it = implementing entmesh
02:26 mikemol richardus: Yeah, I've never implemented a way to effectively transfer entropy from one version of entbuff to another.
02:28 mikemol The thing to do would be to mmap a file for the entropy buffer, including the positional values in header data.
02:49 richardus by data i meant the charting webpage you have and what dates i should be ignoring on it b/c they used the old version
02:50 richardus i'm also surprised to read about your lossyness in yesterday's log.  i guarded my add entropy fcntl with check entropy calls before and after to make sure everything was working right and there was no lossyness at all
03:57 mikemol The old version doesn't show on that chart. I don't think I've run it in over a month.
04:47 mikemol Just pushed a couple minor changes. etools now comes with manpages for entwatch and entbuff, and there's a pre-release checklist. Once I'm through that checklist, the package will be ready for a 1.0 release.
05:21 richardus joined #etools
05:28 mikemol joined #etools
12:54 mikemol richardus: It's possible my assessment of the lossy writes occurred while fwrite() wasn't really working at all for me, but my data was noisy enough that I thought it was.
12:55 mikemol I'll have to try modding entbuff to use while() loops for reading and writing again. That would seriously improve the charge rate.
13:02 mikemol Just killed the entbuff instances on RC, and spawned a single process with parameters of "-i 192 -l 64 -b 2097152".
13:03 mikemol That should result in it charging much faster. It may result in a slight slowdown on the server, though, as ASLR consumes about 128 bits per process spawn, and it's possible for the entropy level to drop as low as 64 bits.
13:03 mikemol Just giving it a try for a day.
16:30 mikemol joined #etools

| Channels | #etools index | Today | | Search | Google Search | Plain-Text | summary