TimToady fixed slurp.t, now says ok 15 - iso-8859-1 makes mojibake correctly 00:28
turned out to be missing an "unsigned" on the char*
down in src/strings/latin1.c 00:29
I don't know why, offhand, passing negative codepoints reverted to interpretation as utf8, which seems bogus 00:30
that seems to imply some kind of multi-pass decoding which would be bad 00:34
or perhaps the semantics we want to check for combining chars at string concat time is being erroneously invoked between each character here 00:35
some NFG assumption that its incoming bufs are utf8 strings maybe 00:37
anyway, forcing the numbers positive seems to protect from whatever the downstream damage is
02:30 FROGGS_ joined 03:27 ingy joined
nwc10 TimToady: you haven't pushed it yet, have you? 07:07
^Yay!
07:41 dalek joined
dalek arVM: 1b9326f | larry++ | src/strings/ops.c:
teach <cntrl> and <print> about Latin-1 controls
10:08
arVM: ee03a5c | larry++ | src/strings/latin1.c:
latin1 is unsigned chars, not signed
MoarVM: eb722b4 | larry++ | / (2 files):
MoarVM: calculate correct offsets for character props
MoarVM:
MoarVM: The offsets were only correct up to about 0x3000. Above that,
MoarVM: span extents were completely wrong. Also, rather than trying to
MoarVM: keep the offset by relative means, all we really need to do is
timotimo "all we really need to do is <transmission cut off>" 10:21
FROGGS hehe 10:23
we'll never know :o)
masak good thing there's a review link at the botto... nooooooooo 10:38
10:46 tgt joined 13:54 cognominal joined
FROGGS I tried to create a serialization test for RT #121319, but I fail gist.github.com/FROGGS/06cdedfd323abddc6790 14:13
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=121319
FROGGS jnthn: do you have time to muse about that ticket? 16:32
jnthn no 17:04
gotta make dinner in a moment
And serialization things often take a bit of time to hunt down :(
FROGGS k 17:07
np, I have another problem to hunt down 17:08
17:16 jnap joined 18:07 tgt joined 19:31 tgt joined 23:52 diakopter joined