| Geth | MoarVM/main: f27687e211 | ab5tract++ (committed using GitHub Web editor) | src/6model/reprs/VMArray.c Leave a VMArray unchanged when pop or unshift is given the wrong register kind (#2078) pop decremented elems before the switch that checks the register kind, and that check throws before zero_slots clears the slot. So nqp::pop_i on an object array died with "pop expected object register" with the last element already gone from the array, while its slot still held the object pointer. VMArray_gc_mark marks only the first elems slots, so a ... (30 more lines) |
09:59 | |
| ShimmerFairy | ab5tract: Hang on a minute, why are you pushing these AI-generated updates to the main branch now, instead of issuing another PR? I don't know for certain what MoarVM's policies on AI or direct pushes to main are, but this seems like something people would like to review first, no? | 10:43 | |
| lizmat | ShimmerFairy: *I* did the merge | 10:44 | |
| and it was a PR | |||
| and I reviewed it | |||
| github.com/MoarVM/MoarVM/pull/2078 | 10:45 | ||
| ShimmerFairy | Oh, I'm sorry. Nothing about Geth's messages or the linekd github page made it clear it was a PR. (I thought the #2078 was a bug report number) | ||
| lizmat | this one was pretty clear (at least to me) | 10:46 | |
| and MoarVM was clearly in error in the behviour | |||
| ShimmerFairy | (also, it appears Geth happened to miss listing this PR earlier, funny enough) | ||
| lizmat | other PRs that ab5tract added yesterday I'm less sure about | ||
| lizmat checks the webhooks | 10:47 | ||
| ShimmerFairy | Yeah, I'm not 100% against AI or anything (more like 50% perhaps?), I just mistakenly thought that they accidentally pushed to main or something, and esp. given the LLM assistance wanted to make sure it didn't slip by unnoticed. | 10:48 | |
| lizmat | well, in this case I thought it made sense, as it was not ab5tract's primary goal | 10:50 | |
| ab5tract | Nope I don’t have write access to moarvm so those are all merges | ||
| lizmat | they're working on a port of MoarVM to Zig and this came out of porting the logic to Zig | 10:51 | |
| ShimmerFairy | Alright, that's good to know in the future. Like I said, just got the wrong impression from how the fix journeyed into main. | ||
| ab5tract | There were two batches so maybe one from the earlier batch was brought in by lizmat | ||
| ShimmerFairy | Also, about that zig work, I haven't heard of it before. Is there a plan to port MoarVM to zig I don't know about, or is it just like a project for an alternate version or something? | 10:52 | |
| ab5tract | Yeah no worries just clarifying. I appreciate the shout out as i would share the same concerns | ||
| lizmat | it's a private project by ab5tract afaik | ||
| ab5tract | ShimmerFairy: just shaking the trees to see what falls out. If all that comes of it is a safer moarvm I’m fine with that | ||
| I find zig code easier to read than C so it’s primarily education motivated | 10:53 | ||
| ShimmerFairy | I'm vaguely aware of zig being a sorta-backwards-compatible C, with the goal of being safer than it, so seems like a good choice for tree-shaking. (AFAIK Rust is considered the champion for memory safety, but is *way* less C-like ofc.) | 10:54 | |
| Geth | MoarVM/main: 2041fa1b69 | ab5tract++ (committed using GitHub Web editor) | src/6model/reprs/VMArray.c Refuse VMArray splices from sources it cannot copy from (#2079) asplice takes the source's element count through the source's own REPR, so any object with elems gets through, but copy_elements then reads the source as a VMArray: for any other REPR it sets the source's REPR data to NULL, skips the fast path and reads slot_type through that NULL. nqp::splice(nqp::list(1, 2, 3), nqp::hash('k', 1), 1, 0) crashed with ... (68 more lines) |
11:13 | |
| MoarVM/main: 5249be7aa7 | ab5tract++ (committed using GitHub Web editor) | src/strings/normalize.c Fix the output arrays of nqp::normalizecodes and nqp::strtocodes (#2083) MVM_unicode_normalize_codepoints builds the normalized code points in a buffer of result_alloc 32-bit elements and then hands that buffer to the output array, but it set the array's ssize to result_pos * sizeof(MVMuint32): the number of elements produced, in bytes. ssize is the capacity in elements, so the array claimed about four times the ... (29 more lines) |
11:35 | ||
|
12:18
lizmat left,
rakkable left,
lizmat joined
12:19
rakkable joined
|
|||
| ab5tract | lizmat++ | 13:57 | |
| For taking a look at these | |||
| lizmat | I've looked at all, but these were the only ones I was confident enough to merge | 13:58 | |
| ab5tract | Yeah I figured that was the case. Much appreciated | 14:02 | |
| lizmat | I hope that MasterDuke17 ShimmerFairy patrickb will find the time to have a look at the others | ||
|
14:49
rakkable joined
14:50
Geth left
16:22
ab5tract left
16:24
ab5tract joined
16:27
Geth joined
19:45
jdv left,
jdv joined
23:41
disbot left,
disbot5 joined
|
|||