Search
Thursday, 19th of April 2018, 18:37:45 UTC
19:03:59
pfdietz2
Bug submitted doing that.
19:11:42
stassats
just received the mail
19:16:21
pfdietz2
Also can generate a failure with the var bound to 2.
19:23:56
stassats
pfdietz2: ok, that's not really a valid bug
19:24:04
stassats
maybe *reoptimize-after-type-check-max* is a bad thing to lower
19:25:07
stassats
pfdietz2: a better thing to try (setf sb-c::*constraint-propagate* nil)
19:53:43
stassats
ok, that's more interesting, but probably doesn't need fixing
19:54:04
stassats
we are relying on constraint propagation there to choose the right types
19:56:26
stassats
pfdietz2: another candidate, (setf sb-c::*check-consistency* t)
19:57:34
stassats
or (setf sb-c::*loop-analyze* nil)
21:02:14
pfdietz2
Not seeing failures from those, but will keep the machinery grinding.
21:02:55
stassats
wait, *check-consistency* doesn't fail
21:03:22
stassats
but sometimes it just produces warnings
21:04:25
stassats
no, ok, it's mostly errors
21:11:22
Xof
*check-consistency* stuff has possibly been ironed out by pfdietz, because the common symptom without it is a function deleting its entire body, then calling it goes directly into the *elsewhere* error-signalling code
21:12:14
stassats
there was some other stuff, like zombie blocks
21:33:02
pfdietz2
I need to do a mass download of all the Common Lisp projects on github. I've been using quicklisp projects for test generation but more is better.
21:33:57
pfdietz2
3885 repos there in CL.
21:34:29
Bike
that might be counting forks. github search kinda blows.
21:35:01
pfdietz2
I'm sure there is much redundancy.
21:41:40
Xof
so, a quick guess about the OS X bug that has got people excited
21:42:01
Xof
the ABI says "the CPU shall be in x87 mode upon entry to a function"
21:43:38
Xof
I remember a time when there was a similarly weird bug like this, because we weren't clearing DF properly
21:43:47
stassats
in my previous analysis, look at the disassembled code, it does xor r13, r13 ... mov rdi, r13 call strstr
21:44:01
stassats
i don't think that's abi, just some bullshit code, but there's no sources
21:45:01
stassats
what is x87 mode, even?
21:45:32
stassats
do we really go into mmx?
21:45:37
stassats
besides, it's in a foreign thread
21:45:48
Xof
we probably do not go into mmx
21:47:10
Xof
well, in any case, I don't have time to actually look for at least a week
21:47:30
stassats
i'm not looking until the next macOS release
21:48:09
stassats
because it smells like a colossal waste of time
21:48:34
stassats
no guarantees that it'll be fixed, but at least some chance
21:53:07
stassats
or is rdi the third argument?
22:11:11
stassats
maybe that zeroing is deliberate, like "uh oh, use after free"
22:23:59
specbot
https://bugs.launchpad.net/bugs/1765533
22:24:21
stassats
ok, that's non trivial to triage
0:46:15
Lord_Nightmare2
** NICK Lord_Nightmare
Friday, 20th of April 2018, 6:37:45 UTC