about summary refs log tree commit diff stats
path: root/mu.arc
Commit message (Collapse)AuthorAgeFilesLines
* 300 - missed a spotKartik K. Agaram2014-11-231-5/+8
| | | | | | Also uncovered a bug in convert-braces, which helps see a simpler close-offset. Instead of all that crap with first incrementing then waiting for counter to return to 0, just wait for counter to go to -1.
* 299 - labels inside functionsKartik K. Agaram2014-11-221-20/+31
|
* 297Kartik K. Agaram2014-11-221-0/+8
|
* 296 - finally, shared a channel between routinesKartik K. Agaram2014-11-221-2/+2
| | | | Was already working, test was just stale. Reassuring.
* 295Kartik K. Agaram2014-11-221-1/+1
| | | | | Too hard to canonicalize routines in traces, with their nested tables. Give up for now.
* 294 - read/write now pass channel by referenceKartik K. Agaram2014-11-221-15/+15
| | | | But handoff test still failing.
* 291 - figured out the channel-handoff problemKartik K. Agaram2014-11-211-3/+8
| | | | | | | | 'read' and 'write' pass in the channel by value, so they block on different *local* variables. Does this kill my plan to pervasively use call-by-value everywhere? No, we might be able to salvage it if channels are the only shared pointers.
* 290 - how was channel-write-block *ever* working?!Kartik K. Agaram2014-11-211-1/+1
| | | | Oh right, because I wasn't using default-scope when checking to wake up.
* 289 - finally back to deadlock detectionKartik K. Agaram2014-11-211-2/+3
|
* 288 - support sleeping while watching a memory locationKartik K. Agaram2014-11-211-1/+2
|
* 287Kartik K. Agaram2014-11-211-19/+21
|
* 286Kartik K. Agaram2014-11-211-5/+7
|
* 285Kartik K. Agaram2014-11-211-2/+1
|
* 284Kartik K. Agaram2014-11-211-9/+9
|
* 283Kartik K. Agaram2014-11-211-20/+20
|
* 282 - scheduler skips ahead when all routines are asleepKartik K. Agaram2014-11-211-26/+32
|
* 281Kartik K. Agaram2014-11-211-26/+21
|
* 280 - waking up sleeping routinesKartik K. Agaram2014-11-211-5/+7
| | | | Tests still hanging at some point.
* 279 - one new test passing, but sleep now hangsKartik K. Agaram2014-11-211-2/+5
|
* 278Kartik K. Agaram2014-11-211-0/+5
|
* 277 - back up, start refactoring the schedulerKartik K. Agaram2014-11-211-32/+55
| | | | | | | | | In trying to share pipes between routines, I realized my scheduler was actually quite brittle. Changing scheduling-interval* shouldn't be required in most tests, and shouldn't change the outcome most of the time. Current state: all scheduler tests fail, but everything else passes.
* 276 - now fork can take any argsKartik K. Agaram2014-11-191-2/+1
| | | | | | | | | No oargs, though. Hopefully we don't need them. Use channels for passing data back. Drawback: channels must all be passed in by value, and their direction isn't obvious. Hard to tell when multiple threads read/write the same channel. Hopefully it's amenable to static analysis.
* 275 - pass args to forkKartik K. Agaram2014-11-191-3/+4
| | | | Only literals for starters.
* 273Kartik K. Agaram2014-11-181-1/+1
|
* 272Kartik K. Agaram2014-11-181-1/+1
|
* 271Kartik K. Agaram2014-11-181-7/+7
|
* 270Kartik K. Agaram2014-11-181-2/+8
|
* 269Kartik K. Agaram2014-11-181-12/+28
| | | | | | Minor cleanup and code comments. I'm starting to feel the need for formatting primitives, so I don't use comments just to provide section headings.
* 267 - more intuitive channel capacityKartik K. Agaram2014-11-151-0/+1
|
* 263Kartik K. Agaram2014-11-081-3/+4
| | | | | | | | | | | | | | | | | I'm trying to think about how to write a test for the race condition, and how to fix it. One thing that's been hard is even remembering where it lies. It's not between wiping the watch and sleeping on it; that's innocuous because the sleep would just immediately wake up. No, the race condition lies between the empty check and the wipe. For the innocuous race we could just create an atomic wipe-and-sleep. But the more serious race requires a lock. If we need a lock anyway, is there any reason to have two watch variables? I'm going to preserve these alternative functions in the code. Alternatives will only ever be called from other alteratives or tests.
* 262Kartik K. Agaram2014-11-081-4/+4
|
* 259 - finally channel read/write can cause routine to sleepKartik K. Agaram2014-11-071-3/+29
|
* 258 - new channel helpers: empty? and full?Kartik K. Agaram2014-11-071-0/+30
|
* 256 - channel pointers can now wrapKartik K. Agaram2014-11-071-9/+25
|
* 254Kartik K. Agaram2014-11-071-0/+1
|
* 253Kartik K. Agaram2014-11-071-17/+17
|
* 252Kartik K. Agaram2014-11-071-3/+3
| | | | Single idiom for setting oargs.
* 251Kartik K. Agaram2014-11-071-5/+8
|
* 250 - experimental support for _ oargKartik K. Agaram2014-11-071-11/+22
| | | | | | | This radically overhauls our assumption that args must always be lists, so we're probably missing things. Where we do, more tests are required. Only important trace change: .traces/dummy-oarg
* 249Kartik K. Agaram2014-11-071-2/+2
| | | | I've been meaning to fix that misleading label for some time now..
* 247Kartik K. Agaram2014-11-071-3/+8
|
* 246Kartik K. Agaram2014-11-071-6/+7
|
* 244Kartik K. Agaram2014-11-061-2/+2
|
* 242Kartik K. Agaram2014-11-061-13/+0
|
* 241 - 'sleep' on a memory locationKartik K. Agaram2014-11-061-5/+18
| | | | I thought we could be oblivious of channels, but that requires more work.
* 240Kartik K. Agaram2014-11-061-6/+6
|
* 239Kartik K. Agaram2014-11-061-2/+2
|
* 238 - avoid deadlock if all jobs are sleepingKartik K. Agaram2014-11-061-2/+9
|
* 237 - rudimentary 'sleep'Kartik K. Agaram2014-11-061-3/+23
|
* 235Kartik K. Agaram2014-11-061-3/+3
|
generated by cgit-pink 1.4.1-2-gfad0 (git 2.36.2.497.gbbea4dcf42) at 2024-11-08 17:36:51 +0000