Da
Show enters and exits. Hide enters and exits.
| 00:00:36 | stepheneb enters the room. | |
| 00:00:44 | lopex_ leaves the room. | |
| 00:02:10 | shoe leaves the room. | |
| 00:04:33 | shoe enters the room. | |
| 00:09:37 | nari enters the room. | |
| 00:12:03 | brynary leaves the room. | |
| 00:18:44 | outerim leaves the room. | |
| 00:26:59 | blowmage leaves the room. | |
| 00:38:53 | macournoyer enters the room. | |
| 00:46:51 | jashmenn leaves the room. | |
| 00:49:25 | antares_ leaves the room. | |
| 00:52:43 | benny leaves the room. | |
| 00:52:48 | cored enters the room. | |
| 00:53:12 | benny enters the room. | |
| 01:00:19 | boyscout | Fixed rb_str_buf_new to set used bytes to 0. - 2a9ee13 - Brian Ford |
| 01:00:19 | boyscout | Cleaned up capi string specs. - 093f08b - Brian Ford |
| 01:00:19 | boyscout | Fixed rb_str_to_str, all capi string specs pass. - fd0912a - Brian Ford |
| 01:04:21 | boyscout | CI: fd0912a success. 1427 files, 6934 examples, 23104 expectations, 0 failures, 0 errors |
| 01:05:04 | brixen | heh http://www.sinatrarb.com/ is funny |
| 01:05:15 | brixen | I think I'll try smoking some with rbx |
| 01:06:24 | brynary enters the room. | |
| 01:12:38 | wmoxam enters the room. | |
| 01:15:27 | yipstar leaves the room. | |
| 01:29:45 | macournoyer leaves the room. | |
| 01:31:19 | bdimcheff enters the room. | |
| 01:32:32 | macournoyer enters the room. | |
| 01:34:58 | rue | Should attempt to run Ebb or Thin |
| 01:35:59 | macournoyer | I doubt Thin will work |
| 01:36:06 | macournoyer | because of EventMachine |
| 01:37:16 | macournoyer | you have rb_thread_select and the like? |
| 01:39:01 | rue | Ah, I did not recall the dependency |
| 01:39:33 | rue | Trying to think of the simplest non-Webrick server |
| 01:40:00 | macournoyer | Mongrel is the one using the most "standard" stuff |
| 01:40:05 | rue | Actually, probably simplest to just run through Rack via [F]CGI |
| 01:40:40 | macournoyer | do you have Fiber? |
| 01:41:07 | macournoyer | http://github.com/ry/flow/tree/master |
| 01:41:34 | rue | Nope, not since switching to native threads |
| 01:44:31 | gnufied leaves the room. | |
| 02:06:52 | ezmob leaves the room. | |
| 02:12:37 | ezmob enters the room. | |
| 02:19:39 | asap18 leaves the room. | |
| 02:27:03 | asap18 enters the room. | |
| 02:30:29 | imajes leaves the room. | |
| 02:31:02 | asap18 leaves the room. | |
| 02:43:20 | brynary leaves the room. | |
| 02:47:25 | binary42 enters the room. | |
| 02:50:08 | gnufied enters the room. | |
| 03:00:16 | tongueroo leaves the room. | |
| 03:08:15 | macournoyer leaves the room. | |
| 03:14:01 | qbproger leaves the room. | |
| 03:14:59 | asap18 enters the room. | |
| 03:26:02 | cored leaves the room. | |
| 03:29:05 | tongueroo enters the room. | |
| 03:29:22 | binary42 leaves the room. | |
| 03:57:55 | tongueroo leaves the room. | |
| 04:00:50 | moconnor enters the room. | |
| 04:35:13 | moconnor leaves the room. | |
| 04:35:19 | antares_ enters the room. | |
| 05:02:11 | jero5 leaves the room. | |
| 05:09:16 | stepheneb leaves the room. | |
| 05:29:31 | mediogre enters the room. | |
| 05:37:07 | asap18 leaves the room. | |
| 06:01:14 | ezmob leaves the room. | |
| 06:08:19 | moconnor enters the room. | |
| 06:09:23 | moconnor leaves the room. | |
| 06:10:23 | bdimcheff leaves the room. | |
| 06:19:27 | boyscout | Fixed up capi hash specs. - b82bb14 - Brian Ford |
| 06:19:27 | boyscout | Added basic capi Hash functions. - 7b84195 - Brian Ford |
| 06:19:49 | brixen | I think I will get mongrel running tomorrow |
| 06:23:57 | boyscout | CI: 7b84195 success. 1427 files, 6934 examples, 23104 expectations, 0 failures, 0 errors |
| 06:28:46 | rue | Need to solve the direct pointer issue, if I recall correctly |
| 06:30:29 | brixen | I will add that tomorrow |
| 06:30:45 | brixen | RSTRING_PTR and RARRAY_PTR |
| 06:31:07 | brixen | also have to fix rbconfig so mkmf loads |
| 06:31:21 | brixen | just imported mkmf from 1.8 head |
| 06:33:27 | rue | Managing the pointers and flushing changes back should not be too bad. Is the hacked rbconfig generator still around? |
| 06:34:58 | vborja leaves the room. | |
| 06:37:38 | rue | drbrain: Did you come across this one? http://www.eecs.harvard.edu/~cduan/technical/git/ |
| 06:42:14 | somebody__ leaves the room. | |
| 07:03:54 | rue | Op, that is not the right one.. http://ftp.newartisans.com/pub/git.from.bottom.up.pdf |
| 07:04:56 | krawek_ leaves the room. | |
| 07:29:48 | evan | brixen: you still awake? |
| 07:30:17 | sunblush leaves the room. | |
| 07:33:30 | rue | Probably not |
| 07:34:22 | evan | seems like not. |
| 07:35:52 | evan | got some awesome insight from John Rose on removing the GIL |
| 07:37:36 | evan | it's nothing ground breaking, but it's how hotspot does it |
| 07:39:00 | rue | Cool, what is the general approach? |
| 07:40:50 | evan | they simplify have a number of internal VM tasks that must be done syncronized |
| 07:40:56 | evan | and have specific mutexs for it |
| 07:41:07 | evan | ie, they do it like you would a normal multithreaded program |
| 07:41:10 | evan | nothing magic |
| 07:41:26 | evan | he advised me on some techniques they use for lock management |
| 07:41:57 | evan | i thought they relied on object locks, but they don't at all |
| 07:42:04 | evan | if you never used object locks, you get weird results |
| 07:42:08 | evan | but the VM would never crash |
| 07:42:22 | evan | i think it's that insight that I find quite interesting |
| 07:42:52 | evan | i kind of thought they'd be using some mythical, uber complicated strategy |
| 07:42:57 | rue | Hm. Which operations are they protecting? |
| 07:43:00 | evan | thankfully, I was wrong. |
| 07:43:13 | evan | oh, things like allocating from the code cache |
| 07:43:17 | evan | or adding new class |
| 07:43:24 | evan | adding new methods |
| 07:43:36 | evan | he told me where to look in openjdk for the list of all the locks used |
| 07:44:49 | rue | Interesting.. rest is user's responsibility, essentially? |
| 07:44:53 | evan | yep |
| 07:45:09 | evan | they do enough to make sure that the VM doesn't corrupt it's data structures |
| 07:45:23 | evan | and that the user can't access memory it doesn't own |
| 07:46:31 | evan | which is quite doable for us |
| 07:46:44 | evan | even in the short term |
| 07:46:56 | evan | it becomes a task of slowly widdling away from using the global lock |
| 07:47:02 | rue | Should be, sure. At that level, the turtles are not getting in the way much |
| 07:47:02 | evan | into using more specific locks around sections |
| 07:47:13 | rue | Semi-random segfault running CGI specs, by the by, immix::Block::from_address is returning NULL in mark_address (although the vm trace is 380 frames deep at that point, too) |
| 07:47:26 | evan | hm |
| 07:47:33 | evan | i'll need at least the top 10 frames to figure that out |
| 07:47:39 | evan | is there a simple repro? |
| 07:51:00 | rue | `bin/mspec run -fs spec/frozen/library/cgi` without .rbcs at least |
| 07:51:05 | rue | Bunch of blocks |
| 07:51:14 | rue | http://gist.github.com/86592 |
| 07:53:53 | evan | hm. |
| 07:54:11 | evan | that likely means that immix tried to mark the block of an address it doesn't own. |
| 07:55:56 | evan | ok, i'll poke at that in the morning. |
| 07:55:59 | evan | i'm going to head to bed. |
| 07:56:09 | rue | Yep |
| 08:00:37 | kronos_vano enters the room. | |
| 08:02:52 | bitsweat leaves the room. | |
| 08:08:28 | imajes enters the room. | |
| 08:08:35 | mutle_ enters the room. | |
| 08:22:53 | SoreGums- leaves the room. | |
| 08:23:02 | SoreGums- enters the room. | |
| 08:24:58 | mutle leaves the room. | |
| 08:52:53 | joachimm_ enters the room. | |
| 08:53:29 | SoreGums- leaves the room. | |
| 08:53:34 | SoreGums- enters the room. | |
| 08:54:30 | naeu enters the room. | |
| 08:58:35 | joachim__ enters the room. | |
| 09:09:12 | joachimm leaves the room. | |
| 09:16:16 | joachimm_ leaves the room. | |
| 09:20:32 | jarib_ enters the room. | |
| 09:23:39 | kronos_vano leaves the room. | |
| 09:23:43 | somebody_ enters the room. | |
| 09:38:44 | dgfitch leaves the room. |