Search
Thursday, 22nd of March 2018, 23:23:17 UTC
23:23:59
dougk_
but what's worse is that we're using "(char*)thread + dynamic_values_bytes" in 25 different places
23:24:03
stassats
does it mean it redoes scavenge_interrupt_contexts?
23:24:39
dougk_
that's why i don't like the contexts appearing as part of TLS. Did that break RISC ? I guess I forgot to check
23:25:08
stassats
i tried not zeroing them, didn't help
23:25:47
dougk_
i think it's that it broke precise GC, which all happen to be risc. I'll look at, this is obviously my doing
23:27:12
stassats
i would've just preprocessed out zeroing on riscs, but if i knew the actual reason it segfaults
23:27:37
dougk_
but you're certain that's the breaking change ?
23:28:00
stassats
the zeroing bit specifically
23:28:32
stassats
and it's not the zeoring segfaulting
23:29:30
dougk_
anyway, the weird thing for me is that our QPX image has 10KB of tenured garbage immediately on restarting an image produced by S-L-A-D (and not letting it even get to main)
23:29:41
stassats
maybe something doing with PINNED_OBJECTS?
23:29:45
dougk_
i was trying to fix that
23:30:18
dougk_
i'll look tonight. i gotta get my kids to bed
23:30:26
dougk_
well, it is tonight. I'll look later tonight
23:31:55
stassats
switched write_TLS(PINNED_OBJECTS, NIL, th); and memset places
23:33:28
dougk_
of course, because reading PINNED_OBJECTS looks at TLS
23:33:45
stassats
it wants a list, not a fixnum
23:34:22
stassats
ok, now to sparc, for some reason it's failing without having threads enabled
23:34:40
stassats
dougk_: is your notinline thingy APPLied?
23:37:09
stassats
lvar-fun-type is supposed to honour :notinline
23:40:57
stassats
huh, (multiple-value-call 'bar (mm)) gets derived, (multiple-value-call #'bar (mm)) doesn't
23:50:57
scymtym
SLIME uses SB-SPROF::CALL-GRAPH-FLAT-NODES which i removed. fix SLIME or put it back in?
23:52:02
stassats
fixing slime is easy, but disseminating a fix is harder
23:52:49
stassats
ok, it doesn't work without it
23:53:19
stassats
scymtym: better to put it back in
23:54:15
stassats
or put whatever in so that M-x slime-sprof-report works
23:58:29
scymtym
stassats: ok, i'm making it an alias for GRAPH-VERTICES (which is unsorted). i can make SLIME sort the vertices properly as a separate change
0:12:10
stassats
arm64 broken with -sb-thread, since no idea when
0:12:13
stassats
can't say i care much
0:23:09
stassats
ok, this is fun, bisecting sparc, a revision doesn't just fail, it spins at 100%
0:23:25
stassats
no idea how to mark it, i guess it's bad too
0:25:14
stassats
and sparc is really slow
0:35:28
stassats
each revision fails differently
0:48:52
stassats
SBCL_MAKE_PARALLEL did something wrong on sparc
0:49:36
stassats
will have to unset, things will get even slower
11:09:10
stassats
bisected sparc to "disassembler: don't note Lisp TLS use if segment override present"
11:09:14
stassats
that's clearly wrong, isn't it?
11:09:51
stassats
is my 'good' revesion even good?
Friday, 23rd of March 2018, 11:23:17 UTC