|
00:21
vendethiel joined
00:47
vendethiel joined
01:47
jimmy__ joined
02:00
vendethiel joined
|
|||
| dalek | arVM/dont-free-callback-strings: cb3d176 | (Rob Hoelz)++ | src/core/nativecall.c: Mark callback Str parameters as not to be freed If a library is passing a string to a callback function, it is very seldom the case that the callback assumes ownership of that string and responsibility for its destruction. Until we add traits to annotate parameters to NativeCall (with regards to memory management), I think it would be best to assume that callback strings should *not* be freed |
02:21 | |
| hoelzro | I made a PR so that others could discuss my change instead of me brashly merging it in | 02:22 | |
| please let me know if this is a terrible idea =) | 02:23 | ||
|
04:51
vendethiel joined
05:22
vendethiel joined
05:45
vendethiel joined
06:50
vendethiel joined
07:14
vendethiel joined
07:36
jimmy_ joined
07:48
FROGGS joined
07:50
zakharyas joined
08:45
vendethiel joined
09:08
vendethiel joined
10:25
jimmy_ joined
11:41
colomon joined
13:40
vendethiel joined
14:49
vendethiel joined
15:31
carlin_ joined
17:15
vendethiel joined
18:07
vendethiel joined
18:24
vendethiel joined
|
|||
| jnthn | hoelzro: Your idea makes sense to me at first blush | 19:42 | |
| I suspect the way it is now is a foolish consistency, and I'd prefer a sane default over that. :) | 19:43 | ||
| (Also, I didn't actually know "free the things" was the current way it was. Probably I shoulda thought about it at some point. :)) | |||
| hoelzro | jnthn: I think it would be best to have traits one can add to specify how memory management should be done, but this should do for now | 20:40 | |
| do you mind if I push that to master, then? | |||
| jnthn | hoelzro: +1 on the traits | 20:42 | |
| hoelzro: +1 on the push to master too; better to get it out there for testing against the module ecosystem | |||
| I think it's the right thing. | |||
| hoelzro | cool, thanks! | ||
| dalek | arVM: cb3d176 | (Rob Hoelz)++ | src/core/nativecall.c: Mark callback Str parameters as not to be freed If a library is passing a string to a callback function, it is very seldom the case that the callback assumes ownership of that string and responsibility for its destruction. Until we add traits to annotate parameters to NativeCall (with regards to memory management), I think it would be best to assume that callback strings should *not* be freed |
20:43 | |
| jnthn | hoelzro: Please can you file an issue in the nqp repo that notes JVM also needs to track this? | 20:44 | |
| hoelzro | jnthn: you got it | ||
| jnthn | Uh, and Parrot too I guess. | ||
| That's likely the eaiser of the two. | |||
| dalek | href="https://moarvm.org:">moarvm.org: d9a2bc8 | jonathan++ | / (3 files): Update site for latest release. Includes homepage and releases history update. Also, the roadmap is given an overhaul to re-focus it on providing things we will need soon in Rakudo, and add many items for the future (without goal dates, since they are not blockers for Rakudo, and can be picked off in the near future or the more distant future as resources are available to do them). |
22:09 | |
| jnthn | We also sound a bit more sure of ourselves on the homepage now, 'cus we are, well, a bit more sure of ourselves. :) | 22:10 | |
| Unless you're on OSX... :P | |||
| lizmat | indeed -( | 22:17 | |
| jnthn will have to get back on the bug hunt with that after the weekend. | 22:18 | ||
| Sleep time; 'night :) | 22:21 | ||
| lizmat | gnight jnthn | 22:22 | |