<durandal_1707>
michaelni: no, i guess it help with crashes
<Lynne>
Daemon404: 1024, the decoder's base algorithmic delay
<durandal_1707>
michaelni: you know that negative linesizes makes big security loophole
<Daemon404>
Lynne, i believe for he it shoul be 2048. or at least that is what is coded in these mp4s.
<Lynne>
in hindsight, it should be 512, because mdct is before SBR
<Daemon404>
(the sgpd box has it)
<michaelni>
durandal_1707, you added a big security loophole in that commit
<Lynne>
disregard seeking for now, it's buggy because lavc/decode is buggy
<Sean_McG>
michaelni o/
<durandal_1707>
michaelni: really? where?
<Daemon404>
Lynne, decode is buggy?
<michaelni>
off by 1 in the check_pixel_ptr() for some of the uses
<durandal_1707>
wiht normal linesizes?
<Lynne>
well, during seeking, like I said, it locks up audio, skips more than it has to, and overall doesn't quite work yet
<michaelni>
durandal_1707, yes
<Daemon404>
ok
<michaelni>
durandal_1707, heap-buffer-overflow and writing at that
<durandal_1707>
can it be really exploited or this is only theory
<Daemon404>
i will apply your patches tomorrow locally to poke and review
<Daemon404>
post-midnght brain reviews are historically a poor choice for m.
<Daemon404>
me*
<michaelni>
i have sample that writes probably whole line out of array, i have no exploit
dellas has joined #ffmpeg-devel
<durandal_1707>
michaelni: with negative linesize it crashes
<durandal_1707>
bigger exploit
<michaelni>
durandal_1707, can you please send your patches to ML instead of direct commits. This off by 1 issue would likely have been caught in review
lexano has quit [Ping timeout: 255 seconds]
<durandal_1707>
michaelni: where is off by one?
<michaelni>
FLI_DTA_COPY / FLI_COPY
<Lynne>
Daemon404: could you look into where to multiply the seek value in mov?
<michaelni>
the loop simply copies one line more now it seems than before
<Lynne>
I tried, but it's a mess, it's used in various places for various uses, and I just know if I do it in the wrong spot, it'll break something else
<Daemon404>
Lynne, we dont even expose the seek value in mov atm
<Daemon404>
so "nowhere"
<Daemon404>
we dont even parse the box atm...
<Daemon404>
i was planning to add that
<Daemon404>
though where to shove that info is unclea.
<Daemon404>
unclear even
<Lynne>
err, sorry, not the seek value, the skip samples value
<Daemon404>
oh that
<Daemon404>
i was mistaken before, i didnt realize it was rescaling to sample rate already
<Daemon404>
so it is in fact already doubling it
<Lynne>
oh...
<Lynne>
the it must be the sbr enhancements
<Lynne>
though I don't get why for aac he_v2 it's more
<Lynne>
and it doesn't matter what type I use for fdkaac -G
<Daemon404>
whether you need to add additional delay to skip_samples depends on whether the program that made the mp4 included sbr in the edit list or not...
<Lynne>
is it getting parsed incorrectly for v2?
<Daemon404>
it should be correct for v1 or 2...
<Daemon404>
it is bigger for fdk for v2 because i think fdk uses more priming for v2
<Daemon404>
3072 samples of priming (?)
<Daemon404>
oh wait no
<Daemon404>
6144 samples?
<Daemon404>
that seems like a lot
<michaelni>
Sean_McG, o/
<Daemon404>
Lynne, anyway... will contiue tmr
<Daemon404>
need sleep
derpydoo has quit [Quit: derpydoo]
<Lynne>
video game codec, video game container
<Lynne>
next year someone will port doom to run on mp4 with aac
<Sean_McG>
LOL
<Sean_McG>
or Bethesda will get ideas and port Skyrim to it
<Lynne>
a janky game on a janky container, how appropriate
noonien88 has joined #ffmpeg-devel
noonien8 has quit [Ping timeout: 272 seconds]
noonien88 is now known as noonien8
dellas has quit [Remote host closed the connection]
navi has quit [Quit: WeeChat 4.0.4]
thilo has quit [Ping timeout: 240 seconds]
thilo has joined #ffmpeg-devel
thilo has quit [Changing host]
thilo has joined #ffmpeg-devel
tufei_ has quit [Ping timeout: 256 seconds]
Sean_McG has quit [Quit: leaving]
lemourin has quit [Read error: Connection reset by peer]
lemourin has joined #ffmpeg-devel
jamrial has quit []
MisterMinister has joined #ffmpeg-devel
Mista_D has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 240 seconds]
MisterMinister has joined #ffmpeg-devel
Mista_D has quit [Ping timeout: 240 seconds]
<pross>
"FFmpeg, brought to you by Carls Jnr."
durandal_1707 has quit [Ping timeout: 255 seconds]
durandal_1707 has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 240 seconds]
kurosu has quit [Quit: Connection closed for inactivity]
qeed has quit [Remote host closed the connection]
qeed has joined #ffmpeg-devel
Traneptora has quit [Quit: Quit]
AbleBacon has quit [Read error: Connection reset by peer]
mkver has joined #ffmpeg-devel
another is now known as another|
Krowl has joined #ffmpeg-devel
dellas has joined #ffmpeg-devel
dellas has quit [Remote host closed the connection]
cone-475 has joined #ffmpeg-devel
<cone-475>
ffmpeg zheng qian master:4dbfb52230c4: doc/decoders: correctly note an option's default in libaribcaption
<cone-475>
ffmpeg zheng qian release/6.1:80daebdfdff4: doc/decoders: correctly note an option's default in libaribcaption
<another|>
durandal_1707: What I give NG, is that your patches could use a bit more explaining in the commit message. Maybe not as much as he asks for, but a bit more than nothing on patch 1.
<another|>
Though, to be clear, you are far from the only sinner in this regard.
<another|>
And with that: Good morning
<durandal_1707>
get a grip
Krowl has quit [Read error: Connection reset by peer]
dellas has joined #ffmpeg-devel
dellas has quit [Remote host closed the connection]
<another|>
on what?
* elenril
slaps durandal_1707 with a large trout
<elenril>
stop being rude to people
<another|>
elenril: Animal cruelty! PETA!!!
ccawley2011 has joined #ffmpeg-devel
<courmisch>
the trout was already dead, so it's ok
<durandal_1707>
elenril: i just do what NG do
Krowl has joined #ffmpeg-devel
<elenril>
he should not be a model for you
<durandal_1707>
my model is elenril
dellas has joined #ffmpeg-devel
navi has joined #ffmpeg-devel
kurosu has joined #ffmpeg-devel
zsoltiv_ has quit [Quit: Left]
zsoltiv_ has joined #ffmpeg-devel
dellas has quit [Remote host closed the connection]
<JEEB>
also for the record, we would have to have a clear hosting responsibility definition. that's kind of why gitlab is probably in lead since we could just utilize the resources videolan already puts into its instance
<elenril>
"just"
<elenril>
I don't think it's quite that simple
<JEEB>
yea I am simplifying it
<Lynne>
I wouldn't trust blender with making good decisions
<Daemon404>
hmm, exposing preroll from the sgpd box is going to be annoying
<Daemon404>
either we need to send new values on every seek, or we need a new way to pass through preroll in the api
<Daemon404>
(this is because the sgpd box stores preroll as "number of packets" and not "number of audio samples")
<Daemon404>
(because it isnalso meant for video)
<Daemon404>
cant work in our current framework except for cases when all packets have the same duration
<Daemon404>
will think about that more later...
lexano has joined #ffmpeg-devel
cone-595 has joined #ffmpeg-devel
<cone-595>
ffmpeg Andreas Rheinhardt master:02064ba3a377: fftools/ffmpeg_mux_init: Restrict disabling automatic copying of metadata
Krowl has quit [Read error: Connection reset by peer]
<jamrial>
mkver: a test for that would be nice
qeed_ has joined #ffmpeg-devel
qeed has quit [Ping timeout: 240 seconds]
Krowl has joined #ffmpeg-devel
<jamrial>
Daemon404: SKIP_SAMPLES type packet side data would solve it i guess, but it's also audio only
mkver has quit [Ping timeout: 260 seconds]
<Daemon404>
jamrial, outputting different side data when seeking vs linear seems mildly evil
<Daemon404>
it would also require every demuxed to implement it
<Daemon404>
rather than just lavc
derpydoo has joined #ffmpeg-devel
<j-b>
good morning folds
<durandal_1707>
folded
<kierank>
folding@home
<Daemon404>
durandal@home
dellas has joined #ffmpeg-devel
Anuj has joined #ffmpeg-devel
<Anuj>
Hey! I'm new to open source and github. Can someone help me setup FFmpeg workspace?
<JEEB>
clone the repository, create a build root (directory) and move to it, then `path/to/ffmpeg/configure` from that build root and see if you are successful. if yes, running `make` should start the build. errors appear? ffbuild/config.log has details in your build root
dellas has quit [Remote host closed the connection]
dellas has joined #ffmpeg-devel
<Anuj>
Okay Sir.
Anuj has quit [Quit: Client closed]
<courmisch>
you know you have a quality build system when the documentation says that parallel build is not supported
Krowl has quit [Read error: Connection reset by peer]
<courmisch>
Using Make as a shell script
<courmisch>
is the purge of MMX to fix ABI violations only?
<courmisch>
cause there are other old SIMD extensions of dubious ongoing and future prospects
cone-595 has quit [Quit: transmission timeout]
<jamrial>
courmisch: not really
<jamrial>
removing all mmx and 3dnow would let us get rid of all the emms calls scattered around
frankplow_ is now known as frankplow
<courmisch>
jamrial: isn't removing EMMS all about fixing ABI violations?
<jamrial>
maybe? the argument i saw in favor of mmx removal is "no one is using a pentium 2 anymore"
<kierank>
it remove the emms calls
<courmisch>
ARMv6 probably doesn't have a whole lot of active users either
<jamrial>
i'm sure there are more phones running that than functioning desktops with 20+ years old cpus
<jamrial>
but yes, emms removal is a good reason too
<kierank>
there's a lot of random emms in there even for non-mmx
<courmisch>
ARMv6 phones running *future* releases of FFmpeg?
<courmisch>
there are barely any ARMv7 phones left
<courmisch>
but but admittedly ARMv6-SIMD does not break ABIs
Krowl has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
Anuj has joined #ffmpeg-devel
<Anuj>
JEEB can you send the above messages again?
<durandal_1707>
Anuj: clone the repository, create a build root (directory) and move to it, then `path/to/ffmpeg/configure` from that build root and see if you are successful. if yes, running `make` should start the build. errors appear? ffbuild/config.log has details in your build root
<Anuj>
durandal_1707 There were more of them?
<durandal_1707>
nope
<Anuj>
But how to check old messages?
Anuj has quit [Quit: Client closed]
AbleBacon has joined #ffmpeg-devel
<Lynne>
elenril: ping on the ff_decode_skip_samples patch
Xaldafax has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
Krowl has joined #ffmpeg-devel
<durandal_1707>
j
Traneptora has joined #ffmpeg-devel
dellas has quit [Remote host closed the connection]
dellas has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
mkver has quit [Remote host closed the connection]
mkver has joined #ffmpeg-devel
mkver has quit [Ping timeout: 255 seconds]
MisterMinister has joined #ffmpeg-devel
iive has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
<jdarnley>
Which CPUs introduced the VPCLMULQDQ cpuid flag? Did it come alongside avx512?
<jdarnley>
Do I have any instrcution references old enough?
<Lynne>
it says VL, so pretty much since the start
<Lynne>
why do we limit image size by stride < int_max?
<Lynne>
oh... why the hell do we have int strides?
<jdarnley>
So many old functions
<Lynne>
I thought we changed that years ago
<jdarnley>
all with int aruments
<jdarnley>
I've not been keeping up
<Lynne>
we should fix that for 7.0
<jamrial>
Lynne: AVFrame has int strides, as do the oldest imgutils functions
<jamrial>
new ones use ptrdiff_t
<jamrial>
i don't know how breaking changing the type in a major soname bump would be, but afaik we normally introduce new fields for that
<jamrial>
but imagine replacing AVFrame->linesize...
dellas has quit [Remote host closed the connection]
dellas has joined #ffmpeg-devel
<Lynne>
11188 occurrences
<Lynne>
*correction. 5268
<Lynne>
of which about 2k go into function calls
l-boogie has joined #ffmpeg-devel
l-boogie has quit [Client Quit]
Krowl has joined #ffmpeg-devel
<Lynne>
Daemon404: could you just send preroll if it's non-zero?
<Lynne>
except for the very first frame
<Lynne>
not sure how we'd apply it if it's sent on every frame
<kierank>
19:30:53 <jdarnley> Which CPUs introduced the VPCLMULQDQ cpuid flag? Did it come alongside avx512?
<cone-943>
ffmpeg Reimar Döffinger master:9dd49c8b52c3: libavutil/log.c: only include valgrind header when used.
<courmisch>
mordorspeak
<Lynne>
at least they call an xor an xor, not nor
<courmisch>
RISC-V calls xor xor
<jdarnley>
kierank: We do but I saw an opportunity
<kierank>
Ok
<kierank>
There is some difference iirc between the avx512 one and normal one
<Gramner>
clmul is a separate cpuid flag. it was added with intel westmere
<Gramner>
I should add it to x86inc. i just never used it myself so never bothered
<Gramner>
and you should check the cpuid flag before using it, there are some weird cpus that doesn't support it despite having fairly modern capabilities otherwise
<jdarnley>
not the base pclmul
<jdarnley>
the vex one
<Gramner>
that one is clmul && avx
<jdarnley>
vex 256 one appears to have its own flag
<courmisch>
somebody claims to have broken RSA 2048 with QC
<Gramner>
the >128-bit one one is quite recent. i think we check for that as part of avx512icl
Krowl has quit [Read error: Connection reset by peer]
<Lynne>
courmisch: "Gerck said all his "QC computations were done in a commercial cellphone, or a commercial Linux desktop," at a capital cost of less than $1,000. "No cryogenics or special materials were used.""
<Lynne>
lk99 was more credible!
<courmisch>
that's the gist of what the cryptography expert in my team said
<Lynne>
once every month or two I go for a dive in the world of quantum computing hype articles
<Lynne>
it's the best, no one has any idea what to say, so they just glue stuff they've heard
<Lynne>
every november or december they get a full injection of hype from IBM, which dries up by may
<courmisch>
Lynne: rewrote opus postfilter, see ML
<Lynne>
wow, that's a huge speedup
<courmisch>
slides don't suck so much on this CPU as the proto
<courmisch>
in fact they seem pretty good, though with latency
MisterMinister has quit [Ping timeout: 255 seconds]
derpydoo has quit [Read error: Connection reset by peer]
ccawley2011 has quit [Read error: Connection reset by peer]
jkhsjdhjs has joined #ffmpeg-devel
cone-943 has quit [Quit: transmission timeout]
mkver has quit [Ping timeout: 240 seconds]
dellas has quit [Remote host closed the connection]
HarshK23 has quit [Quit: Connection closed for inactivity]