|
Post closed tickets in your report. | Note: This channel is for our weekly status meetings (Tuesdays at 19:30 UTC); you probably want #parrot instead. | irclog: irclog.perlgeek.de/ Set by moderator on 29 September 2011. |
|||
|
00:04
alvis joined
01:16
whiteknight joined
05:07
dduncan joined
05:08
dduncan left
09:33
lucian joined
12:49
bluescreen joined
15:40
contingencyplan joined
16:24
bluescreen joined
18:21
dduncan joined
18:22
alvis joined,
dduncan left
19:00
NotFound joined
19:02
whiteknight joined
|
|||
| whiteknight | WHAT I DID: Started a new branch to port alarms to windows. Haven't had a chance to finish it. | 19:03 | |
| WHAT I WILL DO: Hopefully something. I've been too busy for hacking and I'm starting to miss it. | |||
| EOR | |||
| alvis | My pre-post report: (1) I've completed a README_revision.pod (general direction-related stuff) and TODO (task lists) and will push it this afternoon; | 19:09 | |
| (2) Started research on an indexer, which I will write in p5; | 19:10 | ||
| (3) Started a pod, manpage for parrot*; | 19:12 | ||
| (4) Figured out *some*/*more* of how github works. | 19:13 | ||
| That's about it (and I will *try* to get a blog post up sometime this evening). | 19:14 | ||
| NotFound | What I did: | 19:15 | |
| -parrot | |||
| * Testing, minor fixes | |||
| -winxed | |||
| * Fixed diagnosing errors in constant propagation in builtins | |||
| * Some fixes and improvements in builtin constant propagation | |||
| * throw "string" variant now generates throw instead of die | |||
| * Avoid generating 'outer' in inner functions that aren't closures | |||
| * Updated and improved examples and tests | |||
| * Minor fixes and refactors | |||
| -other | |||
| * winxedxx improved, closures starting to work | |||
| * Update and fixed WinxedGtk module | |||
| What I will do: | |||
| No plan | |||
| EOR | |||
| Coke | parrot? I remember hacking on parrot. Hopefully will get back to muddle this week. EOR. | 19:16 | |
|
19:20
dduncan_ joined
19:22
dduncan left
|
|||
| cotto | hello | 19:30 | |
| whiteknight | hello | ||
| NotFound | Hola | ||
| Util | Done: Nil. Still blocking on $WORK. Working on a write-up of bad testing practices in Parrot. .end | ||
| Hello | |||
|
19:31
lucian joined
|
|||
| cotto | How's this week been? I see we have a substantial pull request from mls++. | 19:31 | |
| whiteknight | yes, and I would like to get that merged soon. It cleans up exceptions significantly | 19:32 | |
| NotFound | Has someone tested that with HLLs? I've looked at the changes in exception handling and they may be disruptive. | ||
| whiteknight | I think mls tested it with Rakudo. I think they requested those kinds of changes | 19:33 | |
| we can double-check | |||
| cotto | NotFound: I'd consider that pretty important. | ||
| NotFound: have you looked at the code enough to have an idea of how it'd effect winxed? | 19:34 | ||
| NotFound | cotto: I think it will have zero effect, but I expect to be able to profit from it later. | ||
| cotto | NotFound: great | 19:35 | |
| whiteknight | from what I've seen, I expect it to have zero effect on Jaesop | ||
| cotto | I'm reviewing it now. | ||
| being able to tell github to ignore certain files would be really nice here | 19:36 | ||
| whiteknight | yes, yes it would | ||
| cotto | are there any questions or points for discusion? | 19:39 | |
| whiteknight | I don't think so. It's been a relatively slow week methinks | ||
| NotFound | A note to all: please extra-check not returning NULL in STRING* returning functions, specially in vtable functions. | 19:40 | |
| cotto | NotFound: I noticed that you fixed >= 1 of those. | ||
| NotFound | cotto: I think there are more, in places usually not called from pir. | 19:41 | |
| cotto | NotFound: I'd bet on it. You know what happens when you turn over rocks. | 19:42 | |
| NotFound | I'll try to figure a way to signal assigning NULL to string registers in debug builds. | ||
| cotto | It might even be feasible to write a test that does minimal parsing of C to find instances of that bug. | 19:43 | |
| NotFound | cotto: I thought that my opinions about C parsing were well known ;) | ||
| cotto | fsvo parsing | ||
| The test would need to look at the function definitions and make sure that "return NULL" didn't appear. It wouldn't catch everything, but it'd get the obvious cases. | 19:44 | ||
| NotFound | For the record, it can be summarized as "Don't parse C unless you are a C compiler" | ||
| cotto | In general, yes. In this case, a quick and dirty parser (balanced curlies + simple string matching) has potential to be useful. | 19:46 | |
| Coke | might be worth talking to alester about function annotations to see if we can leverage that to avoid NULLS coming out of vtables. | 19:50 | |
| cotto | It seems like the kind of thing he might be into. | ||
| whiteknight | I'm sure we can use an existing static analyzer tool or the like to find instances, not writing a parser ourselves | ||
| reinventing the wheel, especially in this case, might not be a good use of resources | |||
| cotto | that'd be even better | 19:51 | |
| volunteers are welcome | 19:52 | ||
| NotFound | Take a look at src/pmc/callcontext.pmc 311 - A function declared PARROT_CANNOT_RETURN_NULL that has a return NULL; | ||
| Coke | "whoops!" | 19:53 | |
| cotto | perhaps our annotations aren't being honored as much as we thought they were | 19:54 | |
| NotFound | I think I can be able to trigger some segfault from a hand instantiated CallContext | 19:55 | |
| Here is: winxed -e 'var cc = new ["CallContext"]; string s = cc[1]; s = substr(s, 0, 0); | 19:58 | ||
| ' | |||
| cotto | That's pretty great. | ||
| NotFound | Design decision: should we return STRINGNULL or throw an exception? | 19:59 | |
| cotto | It's charming that the comment before return NULL; says /* exception */ | 20:00 | |
| NotFound | A segfault is a notable exception in the code flow, that's right. | 20:01 | |
| cotto | so the system works, I guess | ||
| except not really | |||
| NotFound | Maybe we can add a new decoration: PARROT_CANNOT_RETURN_NULL___SOMETIMES | 20:02 | |
| cotto | thinking | 20:04 | |
| I'm leaning toward throwing an exception in that case. | 20:06 | ||
| thoughts? | 20:07 | ||
| NotFound | I haven't looked enough at the uses of CallContext from PCC, but I think the exception is the better solution. | 20:08 | |
|
20:08
lucian joined
|
|||
| cotto | NotFound: alright. If you have the tuits to implement it, +1. | 20:08 | |
| NotFound | An assertion in usages from PCC will be fine, but not inside the CallContext PMC, because it can be legitimally called from pir. | 20:09 | |
| cotto | Is there anything else we should discuss? | 20:11 | |
| btw, after testing HLLs with mls++' branch, please note in the pull request that the hll works with the branch. | 20:12 | ||
| Let's call this #ps a wrap. | 20:16 | ||
|
20:19
NotFound left
21:36
bluescreen joined
23:37
whiteknight joined
|
|||