<testuser[m]>
11 extra failures on musl + 1 timeout
<testuser[m]>
the trap message shows up in dmesg
<testuser[m]>
need to check which exact test causes that
<illiliti>
can you attach gdb into it?
<illiliti>
gdb shall catch that sigtrap
<testuser[m]>
yea it does but idk which exact test to look for, since some tests trap on glibc too
<testuser[m]>
wait
<testuser[m]>
ill write a script
<testuser[m]>
o there are some tests where its supposed to die with a sigtrap so idk how tf to differentiate between an actual trap and an intended one
<testuser[m]>
bruh
<illiliti>
i ran chromium with --disable-seccomp-filter-sandbox and it crashes on medium.com almost immediately
<illiliti>
so the bug is not in seccomp if i understand correctly
<testuser[m]>
bruh there are no extra traps in musl tests, same tests trap on glibc and same on musl (intended)
<testuser[m]>
the extra failing tests on musl are realted to filesystem permission stuff
<testuser[m]>
only the ForkInNewPidNamespace.ExitWithSignal seems sus cuz it gets stuck
<illiliti>
i see
<illiliti>
is there a way to enable verbose logs?
<illiliti>
like what assert fails
<testuser[m]>
yea they use gtest so it tells u what value mismatched
<testuser[m]>
but this test just gets stuck so it doesnt know that
<testuser[m]>
ill have to put some printfs
<illiliti>
can you disable that test?
<testuser[m]>
wdym
<testuser[m]>
im running all of them one by one
<illiliti>
ah
<illiliti>
i thought that test prevents other tests from printing logs
<testuser[m]>
btw that threadpoolforeg thing belongs to glib
<testuser[m]>
the one that keeps segfaulting
<testuser[m]>
hmm no
<testuser[m]>
bruh im trying to build with symbol_level set but the obj files dont seem to have any debug info
<testuser[m]>
ig ill have to pass -g manually
<illiliti>
what does is_debug do?
<illiliti>
also set this: dcheck_always_on = true
<testuser[m]>
non debug build just enables more optimization
<testuser[m]>
nothing with symbols
<illiliti>
btw you should use -ggdb rather than -g
<illiliti>
or even -ggdb3, since chromium is highly polluted with macros
<testuser[m]>
it'll probably take a whole day to build with that
<illiliti>
yeah
<illiliti>
whole week
<testuser[m]>
https://termbin.com/serb i checked the args and symbol_level does seem to be set properly, no idea how tf they're passing it to the compiler..
<testuser[m]>
nothing in ninja file or in /proc/pid/cmdline of gcc
<illiliti>
have you set is_debug=true?
<testuser[m]>
doesn't seem to be, it's not present in alpine's debug build file either
<testuser[m]>
They just set symbol level
<testuser[m]>
Seem to be required*
<illiliti>
blink_symbol_level=2
<testuser[m]>
Yeah
<illiliti>
v8_symbol_level=2
<testuser[m]>
Both that and symbol_level without blink_ prefix
<testuser[m]>
i took a snapshot of this tree, lets see
<testuser[m]>
[584/9304]
<testuser[m]>
no
<testuser[m]>
1/3rd of the full build
<testuser[m]>
no
illiliti has quit [Read error: Connection reset by peer]
illiliti_ has joined #kisslinux
<illiliti_>
how many cores do you have?
<testuser[m]>
6
<testuser[m]>
Tbh our time would be better spent figuring out broken ipv6 in Firefox
<illiliti_>
i agree
<illiliti_>
but let's deal with chromium first
<testuser[m]>
I grepped for DCHECK in sandbox/linux there's barely 5-10 of them so if more dchecks get triggered after rebuild im just gonna rebuild with dcheck disabled
ejjdhfjsu has quit [Ping timeout: 260 seconds]
illiliti_ has quit [Quit: leaving]
<testuser[m]>
Fuck i could've just ld preloaded the abort function