<geist>
it's very stereotypically JRPG, with all the usual silly plot twists, but it is primarily story driven. got kinda a final fantasy X sort of feel in terms of being fairly linear and everyone has some back story, etc
<kazinsal>
original plan for the day: work on dumb little XT unix clone
<klys_>
mmu or nommu?
<kazinsal>
actual things accomplished today: finding a quiet little beach on the river and enjoying some sunlight
<geist>
haha yeah, i was feeling a little cooped up on the house after having gone to comic con yesterday
<geist>
so went to a local brewery for a while, then took a walk, and now hanging out at starbux. weather is fantastic
<klys_>
san diego?
<geist>
i always get a little bit of a mood slump just after doing something fun (like a vacation or a comic con or whatnot)
<geist>
nah, the seattle one. Emerald City Comic Con. last 4 days. not as big as the san diego one but fun anyway
<klys_>
:)
<klys_>
yes agreed the weather is verynice
<kazinsal>
dang, totally forgot that was this weekend. shoulda used it as an excuse for a road trip haha
<geist>
ah i should have mentioned it ahead of time, but i have a general policy of not announcing trips or whatnot ahead of time on social/public media
<geist>
just a general why not level of paranoia re: folks knowing if i'm around the house or not
<heat>
that's sensible
MiningMarsh has joined #osdev
[itchyjunk] has quit [Ping timeout: 256 seconds]
[itchyjunk] has joined #osdev
spikeheron has quit [Quit: WeeChat 3.6]
spikeheron has joined #osdev
ghee has quit [Quit: EOF]
scoobydoo has quit [Ping timeout: 256 seconds]
zaquest has quit [Remote host closed the connection]
zaquest has joined #osdev
* alpha2023
waves hi
<alpha2023>
Silly question, has anyone noticed hyper-V being... weird(?) Compared to other virtualization tech?
qubasa has quit [Ping timeout: 268 seconds]
<alpha2023>
I've been trying to validate my kernel works on more than qemu, seems fine on baremetal ryzen, virtualbox, parallels on a mbp, but hyperV just runs dog slow
<alpha2023>
And setting virtualbox's emulation to hyper-v on windows shows same results (legacy is fine though)
<zid>
mmmm vmexitty
<heat>
virtualbox's hyperv backend is dog slow
<heat>
at least it used to be
<heat>
what's slow exactly?
Reinhilde is now known as MelMalik
xenos1984 has quit [Read error: Connection reset by peer]
<alpha2023>
It's like. I can see the individual text characters from my kernel logging function being written to the screen like I was at 1200 baud
<heat>
what happens if you try something like serial?
<alpha2023>
Also, I have a small spinner to test my 8254 PIC code which does a "|/-\" char loop at a rate of 1Hz for a bit. It runs like. 10 times slower than it should
[itchyjunk] has quit [Read error: Connection reset by peer]
<alpha2023>
My uart drivers aren't currently working so would be hard to test
<heat>
have you considered that maybe your timing is screwed?
<alpha2023>
I considered it, but it works on three other virtual platforms, as well as on two real computers baremetal boot
<heat>
doesn't matter
<alpha2023>
How you figure?
<heat>
virtualbox's setting is just a paravirtualization thing
<heat>
you could set it to KVM in windows ezpz
<heat>
it's not going to use KVM in windows as you may guess
<alpha2023>
Well. It works fine with KVM qemu on my linux box
<alpha2023>
It's specifically only broken when utilizing hyper-V
<heat>
unless you actually went to the effort of enabling hyperv such that virtualbox needs to use hyperv, it's not going to
<heat>
are you using any kvm paravirt?
<heat>
(in your kernel)
<alpha2023>
Not a lick
<heat>
that's weird
<alpha2023>
It's just very strange behavior
<heat>
so you get a huge difference in performance by changing the paravirtualization dropdown from legacy to hyper-v?
<alpha2023>
Yeah. Legacy runs as expected. Default is slow, Hyper-V is slow (same as default)
<alpha2023>
It works as expected on that box when kernel is booted baremetal via usb
<kazinsal>
iirc the Hyper-V paravirtualization setting passes through a bunch of Hyper-V's actual functions to the guest VM and Hyper-V is kind of not that good at anything that isn't Windows or Linux
gog has quit [Ping timeout: 268 seconds]
<alpha2023>
Ah that's helpful to know
<alpha2023>
So theoretically at least, it's just hyper-V being terrible
<kazinsal>
yeah, it's not great
<kazinsal>
I've done a bunch of enterprise migrations from Hyper-V to VMware because Hyper-V is just a mess
<alpha2023>
*nodnod*
heat has quit [Ping timeout: 256 seconds]
xenos1984 has joined #osdev
scoobydoo has joined #osdev
opal has quit [Remote host closed the connection]
opal has joined #osdev
<klange>
alright, generic glyph cache implemented... calculator app is now speedy even when it wants to draw lots of numbers
<zid>
cool cool
<zid>
when kerning
<klange>
Really doesn't matter for the fonts I have on hand, and I should probably at least improve this glyph cache system to support subpixel x positioning.
<zid>
3x rendering hype
<klange>
3x and integral might be enough, though the main point there is to support subpixel antialiasing, which I refuse to do.
<zid>
You need it for proper kerning though
<zid>
kerning info is more precise than 1 pixel at typical font sizes
<klange>
You need full detail x-advance.
<zid>
(unless the text is beeeg)
<zid>
The AA just comes as a sort of side-effect
<zid>
of having accurate kerning
<klange>
My "here's a string, draw it" thing supports subpixel x-advance, it turns the whole string into a vector description and paints that.
<zid>
yea, this is only relevent once glyph caches become involved, otherwise you can just use the kerning info along with the vectors in realtime
<klange>
I wonder if half-pixel is enough precision... or should I do quarters... either way would happily screw with anyone wanting to do subpixel AA :3
<zid>
(so three m glyphs, at 0, 0.33 and 0.66 pixel offsets, and if your kerning says 0.5 you interpolate between 0 and 0.33)
<zid>
err 0.33 and 0.66 I am good at number
zaquest has quit [Ping timeout: 248 seconds]
<moon-child>
no polynomial interpolation?
catern has quit [Ping timeout: 255 seconds]
elderK has joined #osdev
the_lanetly_052 has joined #osdev
the_lanetly_052_ has joined #osdev
the_lanetly_052 has quit [Ping timeout: 256 seconds]
<mrvn>
My experience is that for a console the font is too small and all that advanced rednering just makes the font blury and have colored shadows. (but then I don't generally have >= 4k display)
<mrvn>
zid: the green frame looks blury to me and I guess the red one has bad spacing between letters.
<moon-child>
Griwes: somebody conjectured that sqrt(x) on gpus is commonly implemented as x*rsqrt(x). Is that true, or is there specialised circuitry for it?
<moon-child>
(also sorry to whack you with gpu-related questions :P)
[itchyjunk] has joined #osdev
catern has joined #osdev
<mrvn>
is rsqrt so much easier to compute that they only have that in hardware?
<moon-child>
presumably it's more important
<Mutabah>
it's a very common calculation in vector maths, so you'd want it to be in hardware if you could
vdamewood has joined #osdev
<kazinsal>
there's a reason Q_rsqrt() { // what the fuck? } is a memorable function, and it's not just because of the evil floating point bit-level hacking
foudfou has quit [Remote host closed the connection]
foudfou has joined #osdev
vinleod has joined #osdev
vdamewood has quit [Killed (calcium.libera.chat (Nickname regained by services))]
vinleod is now known as vdamewood
<mrvn>
but that's an approximation used in games (and for float). I would expect GPUs to be accurate
<Mutabah>
I assume implementing rsqrt and sqrt would take the same amount of die area, so if you have to pick one (and you're a GPU) you'd pick the former
foudfou has quit [Remote host closed the connection]
<mrvn>
You do need x/|x| a lot in graphics.
foudfou has joined #osdev
gog has joined #osdev
[itchyjunk] has quit [Remote host closed the connection]
elderK has quit [Quit: Connection closed for inactivity]
<moon-child>
suppose your rsqrt were fully accurate. Then computing sqrt as x*rsqrt(x) would be fairly accurate, just losing a few bits
<moon-child>
iirc ieee-754 only guarantees exact results for a few primitive arithmetic operations
foudfou has quit [Remote host closed the connection]
vinleod has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
sikkiladho has joined #osdev
gelatram has joined #osdev
Ali_A has joined #osdev
gareppa has joined #osdev
ghee has joined #osdev
eroux_ has joined #osdev
gareppa has quit [Quit: Leaving]
eroux has quit [Ping timeout: 256 seconds]
jimbzy has joined #osdev
tomaw has quit [Quit: Quitting]
SGautam has joined #osdev
tomaw has joined #osdev
nyah has joined #osdev
joe9 has quit [Remote host closed the connection]
eroux has joined #osdev
eroux_ has quit [Ping timeout: 252 seconds]
gelatram has quit [Ping timeout: 252 seconds]
eroux has quit [Ping timeout: 252 seconds]
gelatram has joined #osdev
gelatram has quit [Ping timeout: 252 seconds]
dude12312414 has quit [Remote host closed the connection]
dude12312414 has joined #osdev
sikkiladho has quit [Quit: Connection closed for inactivity]
dude12312414 has quit [Remote host closed the connection]
dude12312414 has joined #osdev
scoobydoob has joined #osdev
scoobydoo has quit [Ping timeout: 248 seconds]
scoobydoob is now known as scoobydoo
the_lanetly_052 has quit [Ping timeout: 256 seconds]
eroux has joined #osdev
eroux_ has joined #osdev
eroux has quit [Ping timeout: 268 seconds]
FreeFull has joined #osdev
Ali_A has quit [Quit: Connection closed]
<LittleFox>
yay, figured out my newlib problem from a few days ago
<LittleFox>
but, of course, there is a new problem: current newlib master seems to be broken, not filling _newlib_version.h with infos
<zid>
crudely fill it in with crayons
gildasio has quit [Remote host closed the connection]
foudfou_ has quit [Remote host closed the connection]
dude12312414 has quit [Read error: Connection reset by peer]
gildasio has joined #osdev
foudfou has joined #osdev
dude12312414 has joined #osdev
dude12312414 has quit [Remote host closed the connection]
dude12312414 has joined #osdev
<geist>
yeah newlib is kinda not fantastic in my experience
<LittleFox>
zid ^^'
joe9 has joined #osdev
<geist>
i have a side project that uses it and IMO it's only really useful for embedded stuff. however it does have releases though, so it may be that they dont particularly expect mainline to be stable
<geist>
or... you could roll a fix
<LittleFox>
they rebuilt their build system
<LittleFox>
and ... I hate autoconf a lot, was fighting it already while rebasing my branch from 4.1.0 to master
<LittleFox>
and tried to fight some more to fix this problem
<LittleFox>
but
<LittleFox>
nope
<LittleFox>
enough autotools for one day
<geist>
ah
<zid>
I want that embroidered on a pillow
<geist>
haha
<LittleFox>
like, they depend on a "feature" of automake since some time this year
<LittleFox>
but automake changed that "feature" two commits after the release used the last time for newlib
<LittleFox>
-.-
<LittleFox>
just hate that whole autotools ecosystem
<LittleFox>
it's fragile
<mjg>
autotools hates you
<GeDaMo>
It's not personal, Autotools hates everyone :P
<geist>
not a fan, i always avoid them. *usually* for bare metal stuff i've avoided it because you rarely need to know too much about the host, except where the tools are
<geist>
it's mostly a different problem set vs compiling thing for various host OSes
<Bitweasil>
Yeah, my bare metal builds tend to end up being a "build.sh" script that just calls the assorted things.
<LittleFox>
a friend just wrote "I am partially convinced that automake was developed as a parody of bad tools and people didn't notice and ran with it"
<Bitweasil>
If I'm being fancy, I'll have a bit that checks for arm-none-eabi-gcc-whatever, but I mostly don't bother.
<Bitweasil>
You don't have to be *good* to get popular.
<Bitweasil>
You just have to be *better than the alternatives.*
<Bitweasil>
And autoconf/etc was certainly that.
<LittleFox>
yeah, was
<geist>
also think it came out of an era when there were more active flavors of unix around
<zid>
Basically all autoconf does is waste about 8 hours of my day to tell me that -lSDL won't be found at link time, if I were to hit make
<zid>
which isn't true because I use a package manager to install anything that comes with autoconf
<Bitweasil>
Right, if you're wanting to make something build on IRIX, Solaris, A/UX, Linux, etc... it's harder.
<mjg>
you don't have to be better either
<mjg>
see php ))
<Bitweasil>
Hey, PHP did some very nice stuff back in the day.
<Bitweasil>
Now, doesn't mean it should have lived anywhere near as long as it did.
<bslsk05>
newsreleases.sandia.gov: Sandia National Laboratories: News Releases : High-speed fire footage reveals key insights for power plant safety
<zid>
insurance is going to be like "nope" if I try to blame electrical fires on why the house burnt down while I was playing with accelerants
<LittleFox>
zid wrong insurance level then
<LittleFox>
:D
<gog>
you weren't playing with acceleraants, your cat was
<zid>
can I get insurance for playing with deoderant like a flame thrower
<gog>
how can a cat know that that's dangerous
bauen1 has joined #osdev
<LittleFox>
I'm insured for up to gross negligence
<LittleFox>
as long as I didn't do something on purpose, it's fine >:]
<LittleFox>
never tried that tho x)
<zid>
gross neglience is 144 counts of negligence right? I'm sure I could manage 145
<LittleFox>
x)
<gog>
a baker's gross
<gog>
the 145th one is just for you :D
<zid>
I prefer a bad lover's gross
<zid>
It's 68, like a 69 but they still owe you yours
<gog>
hah
<gog>
nice
kof123 has quit [Ping timeout: 268 seconds]
foudfou has quit [Quit: Bye]
foudfou has joined #osdev
sortie has quit [Ping timeout: 248 seconds]
sortie has joined #osdev
GeDaMo has quit [Quit: A program is just a bunch of functions in a trenchcoat.]
ripmalware has quit [Remote host closed the connection]
ripmalware_ has joined #osdev
<jjuran>
Sometimes I like to lie down on the grass and watch the butts go by
<klange>
touch grass, see ass
* kazinsal
nods sagely
<dzwdz>
when implementing tcp, can i just ignore the options fields?
<dzwdz>
i can't get linux to respond to my syn-ack
<heat>
ignore?
<dzwdz>
not parse them in any way
<heat>
depends
<heat>
i think so
<dzwdz>
the few ones i looked at seem to only matter if both sides send them
<heat>
if you send a bunch of options on syn and the syn-ack doesn't reciprocate the options aren't gonna affect anything IIRC
<heat>
yeah
<heat>
are you sending the SYN or the syn ack?
<dzwdz>
i'm sending the syn-ack
<heat>
wanna send a pcap this way?
<heat>
i can take a look
<dzwdz>
sure, wait a sec
<zid>
https://hpd.gasmi.net/ Just wanna drop this like while it's vaguely topical, really handy ass website :P
<bslsk05>
hpd.gasmi.net: Hex Packet Decoder
<zid>
I assume it's just pcap underneath
<zid>
or whatever the parser lib is that wireshark uses for pcap
<heat>
ooh
<heat>
cool
<heat>
"This site is powered by Wireshark. "
<heat>
it is literally the thing
<zid>
yea I doubt they're just spawning gui wiresharks in a vm though, they've probably (I hope) ripped out whatever libpacket or whatever is in wireshark's source
<dzwdz>
oops, i think i've figured it out - i loaded the pcap in wireshark just to make sure, turns out i got the mac adresses wrong
<dzwdz>
didn't see that in tcpdump
<zid>
that'd do it
<zid>
Mine is 00:0c:29:7f:d9:71 if you want a different one (note, may not be the one you wanted)
<heat>
zid, wireshark cli?
<zid>
seems a bit heavy, but maybe
<heat>
dzwdz, there you go, rubber ducky debugging
<dzwdz>
^^
<dzwdz>
i wouldn't have seen it if you didn't ask for the pcap though
<zid>
bugs are almost always
<heat>
you should have the pcaps wired up at all times
<heat>
it's soooooooo useful here
<zid>
"Which thing are you overlooking because you thought you already ruled it out"
<zid>
That's why people who refuse to give actual pastebins and stuff are so annoying, cus you just bloody know that the bug is that they're lying to themselves, and thus you
<heat>
can you ping me in some hours once you've got a PR up? kthx
<heat>
I also have an extra special issue "make filesystem writes stable"
<heat>
should be fairly ez
<heat>
it requires you to rework the block device writeback logic and then hunt for bugs in fs/ext2 code
<dzwdz>
are you sure getting me to mess with your os is a good idea
<heat>
it's either me or you
<heat>
so yes, it is a good idea
<dzwdz>
up until a few hours earlier today each packet had a 1/65536 of hanging the machine
<dzwdz>
which was a pretty fun feature
<jjuran>
Never tell me the odds
<heat>
#morning-motivation #grindset
<dzwdz>
actually probably less
<heat>
if you slightly stress test my filesystem code you can get it to 1) get you a corrupted ext2 fs; 2) panic on block writeback bugs; 3) both? 4) something else I probably haven't figured out yet
<heat>
works on tmpfs though!
<heat>
you should join me and jjuran in the illustrious list of Onyx contributors
<zid>
why are you guys here though and not madly refreshing nyaa for honzuki p4 v9 prepub 6
<heat>
yes
<jjuran>
heat: Please remind me what I contributed
<zid>
headache
<jjuran>
:-(
<heat>
you fixed 2 typos in the readme 5 years ago
<zid>
oh nice
<zid>
can I fix factural errors
<heat>
the nvidia open-gpu-kernel-modules contribution before open-gpu-kernel-modules
<jjuran>
how leet :-)
<heat>
you can put "Operating systems expert" on your resume now
<jjuran>
Speaking of which, I contributed a type fix to Perl which became change #31337, I kid you not
<jjuran>
*typo
<zid>
Pedro falcato is a fairly cool name
<heat>
kinda boring really
<heat>
wish I was named Josh
<heat>
really has the big dick in "big dick energy"
<zid>
where bookworm :( It is laaate
<mjg>
jjuran: nice
<heat>
hello matthew garrett of linux kernel fame
<heat>
what brings you here
<mjg>
well hi joel spolsky of stackoverflow fame
<zid>
does anyone have a j-novel club subscription
<mjg>
just wanted to say hi
<zid>
mateusz guzik is a name also
<mjg>
does not have a ring to it, does it
<zid>
I have no idea, I can't read it
<zid>
It looks czech or something
<mjg>
google translate can
<mjg>
OH that's offensive
<zid>
poles hate czech?
SGautam has quit [Quit: Connection closed for inactivity]
<mats1>
slav on slav violence
<mjg>
here is an example
<dzwdz>
it sounds more polish than czech
<mjg>
i talk to a dude at red hat brno office
<zid>
gtrans says 'mateiuz gujjik'
<mjg>
plenty of czech of people as you cna imagine
<mjg>
but some slovaks as well
<heat>
eastern europeans hate other eastern europeans
<zid>
My name is Dick Dickinson
<mjg>
i ask one guy if he is czech and he was like "hell no"
<mats1>
rip pan slavism
<mjg>
zid: i'm mike moneybags
<heat>
rip my man yugoslavia
<zid>
buy me a pony?
<mjg>
were you good?
<mjg>
i can send you some scam nft
<heat>
Peter Jonhson is soooo vanilla
<mjg>
dick dollars
<heat>
not "crummy european dude" enough for effective operating systems development
kof123 has joined #osdev
<mjg>
dick dollars is a great name
<heat>
rusty kuntz is better
<mjg>
i tried to create an account on lichess.org with it and got rejected
<mjg>
:[
<zid>
*withdrawl symptoms*
<mjg>
thus i'm not playing chess
<zid>
pls pirates, upload faster
<mjg>
zid: what are you pirating, now that i would tell on you
<heat>
the best part of not playing chess is that you don't get to play chess
<zid>
Ascendance of a Bookworm Part 04 Volume 09 PREPUB [6/8]
<zid>
it came out 26 entire minutes ago and it isn't up yet
<mjg>
sounded like an anime... and it is
<mjg>
kthxno
<zid>
it's a book
<mats1>
try unicode characters in the account name
<zid>
of which there is a relatively low budget but passable anime adaptation of volume 1-3
<heat>
that's called and all but
<heat>
have you considered reading "Firmware Development: A Guide to Specialized Systemic Knowledge"
<heat>
i stopped playing chess a year and a half ago
<heat>
best decision I've ever made
<zid>
I've only really played a tiny bit
<heat>
1) cringe 2) nerd 3) bad for my heart
<heat>
4) if flags are visible, xenophobia is increased by 400%
<mjg>
:))
<mjg>
my biggest problems was with people would just run the clock down
<klange>
heat: do you have instructions for running Onyx? Trying to use one of the CI build outputs, but it just complains about not being able to read an MADT.
<mats1>
chess is gay
<zid>
Have you considered having an MADT?
<zid>
-machine q35
<heat>
klange, crap
<zid>
or did heat break it
<heat>
i broke it I guess
<zid>
oopsie
<klange>
all those tests and you broke it anyway ;)
<heat>
klange, did you use a nightly?
<zid>
heat accidentally used coffee inhalation instead of CI
<heat>
it should be fixed in normal builds at least
<klange>
This is the latest CI run output :)
<klange>
The check was green!
<heat>
fuckin what
<zid>
It's up!
<zid>
bye nerds
<geist>
uh what's going on?
<heat>
i broke my CI last night
<heat>
wbu
<geist>
ah
<heat>
OH
<heat>
shoot
<klange>
I used to actually run QEMU on CI... I should do that again...
<heat>
yeah I should too
<geist>
yah i need to do the same thing. dreading writing a whole thing to read test results out of the run/etc
<heat>
+1
<geist>
but really should just do it in some python script or whatnot. cheezy but a good first step
<klange>
I have a fun CI issue that comes up every so often where the FAT builder will fail to produce an image because the thing that tries to calculate an appropriate size underestimates, and this doesn't result in Make failing, so the build 'passes' but you end up with an empty CD (ISO contents mirror the embedded FAT)
<heat>
basically what you saw there was a kernel that was compiled without CONFIG_ACPI because it was using the wrong kernel config and yesterday I added that option
<klange>
Ah, classic.
<heat>
x86 is not really usable without CONFIG_ACPI
<heat>
because why would it
<heat>
but... other architectures should be
<heat>
basically if you wanna run Onyx check back in a few hours
<heat>
probably less
<heat>
get a nightly's disk-image.img, run it WITH EFI AND AHCI
<heat>
I haven't found a way to create disk images like those without hacky stuff like this
<heat>
also, Haswell+
<heat>
the .iso's aren't really usable for people because that's what I use for testing so they expect a hard drive with a root partition that they mount
<klange>
Is the disk image bootable on its own?
<heat>
yes, on efi and ahci
<heat>
and it has the full set of packages
<heat>
I'm not sure how I would fit all that crap into the .iso
<heat>
since that would mean they would get put in RAM
* heat
wants squashfs for compression
wxwisiasdf has joined #osdev
<wxwisiasdf>
Hello
<wxwisiasdf>
I've been fiddling with gcc .md files for a while (i have my reasons), however gcc ends up recompiling half of the planet earth even if i change 1 line, why??
wxwisiasdf has quit [Quit: Lost terminal]
<heat>
thats it
xenos1984 has quit [Read error: Connection reset by peer]
<heat>
im working on storage tonight
<heat>
gonna make it async and then fix writeback
<klange>
what keyboard layout is this, portuguese?
<klange>
I should... actually implement layout configuration so I'm not being hypocritical when I ask how do I switch it for my stupid American keyboard [or even my JIS keyboard, which is at least closer to ISO]
dude12312414 has quit [Quit: THE RAM IS TOO DAMN HIGH]
<heat>
funny how that issue affected only that specific configuration
<heat>
GCC + LTO + ubsan enabled
<heat>
anything else hid it
<heat>
this is *exactly* why I need to setup running Onyx under qemu in CI
<heat>
I can't test every config
<klange>
This is why I just only support things I can test - and why I'm quite happy to have my HVF setup on my M1 Air to test aarch64 :)
<heat>
well, I can totally test it
<heat>
but i.e I don't develop with LTO for obvious reasons, and I'm usually using clang and not gcc
<klange>
Ah, my production setup is identical to my dev setup. While I usually let the CI build the final images for releases, sometimes they're manual uploads of whatever my local build spit out :)
<klange>
Not that this is without issue... I strip binaries, for example, and am often bit by it - I have a debugger now, and it's often useless because it can only find dynamic symbols!
<heat>
yeah. in my case I support like 10 or 15 configuration options + two toolchains + three architectures
<heat>
oh yeah totally
<heat>
my case was super cursed
<heat>
since it was LTO, everything was flattened down to a pulp
<heat>
I took a look at the faulting address and realized that it was in the middle of stack space, and I was memcpy'ing to the stack, so the length was probably screwed
<heat>
so I noticed that I never initialized it
<heat>
disappointed that neither clang nor gcc catched that
<klange>
I had a similar thing recently, thankfully in a userspace tool and I could run it under Valgrind in Linux which was happy to report the unitialized access.
<clever>
klange: i saw a blog post once, where the crazy dude converted a windows kernel driver into a linux elf .so file, mocked out all of the dynamic imports, and then ran it under regular old gdb
<klange>
I've been trying to do more of that - building my shit for Linux. Helps to make sure I don't have the wrong ideas about some common APIs.
<clever>
he didnt build it for linux though, he took a pre-built windows driver