michaelni changed the topic of #ffmpeg-devel to: Welcome to the FFmpeg development channel | Questions about using FFmpeg or developing with libav* libs should be asked in #ffmpeg | This channel is publicly logged | FFmpeg 7.1.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
<kasper93>
BtbN: Can you add information about this in comment? So in few years, we still know why it is how it is
<BtbN>
yeah, but not today anymore
<kasper93>
thanks and no hurry
IndecisiveTurtle has quit [Ping timeout: 265 seconds]
IndecisiveTurtle has joined #ffmpeg-devel
minimal has quit [Quit: Leaving]
TheVibeCoder has quit [Ping timeout: 244 seconds]
TheVibeCoder has joined #ffmpeg-devel
IndecisiveTurtle has quit [Ping timeout: 265 seconds]
iive has quit [Quit: They came for me...]
MisterMinister has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
philipl has quit [Quit: leaving]
philipl has joined #ffmpeg-devel
mkver has quit [Ping timeout: 252 seconds]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
Martchus_ has joined #ffmpeg-devel
Martchus has quit [Ping timeout: 245 seconds]
jamrial has quit []
System_Error has quit [Remote host closed the connection]
kasper93 has quit [Ping timeout: 276 seconds]
marth64_ has joined #ffmpeg-devel
Marth64[m] has quit [Ping timeout: 252 seconds]
zip6como has quit [Read error: Connection reset by peer]
bwu25 has joined #ffmpeg-devel
bwu25 has quit [Ping timeout: 260 seconds]
bwu25 has joined #ffmpeg-devel
derpydoo has joined #ffmpeg-devel
bwu25 has quit [Ping timeout: 248 seconds]
<fflogger>
[newticket] lp24db: Ticket #11654 ([undetermined] addtional sps/pps gets added when remuxing mxf files) created https://trac.ffmpeg.org/ticket/11654
<TheVibeCoder>
shouldn't this macro check ptr instead of off
<TheVibeCoder>
or at least both of them?
<TheVibeCoder>
it was added in 2021.
<TheVibeCoder>
no point in adding offset to NULL, but if that happens game is already lost
<JEEB>
offset to nullptr is a case of ACHTUNG
<JEEB>
(might be worth an assert, even)
<TheVibeCoder>
shhh, do not make ffmpeg code secure like rust one
<JEEB>
;)
derpydoo has quit [Quit: derpydoo]
IndecisiveTurtle has joined #ffmpeg-devel
Painter has joined #ffmpeg-devel
Anthony_ZO has joined #ffmpeg-devel
kasper93 has joined #ffmpeg-devel
Painter has quit [Excess Flood]
mkver has joined #ffmpeg-devel
a1rwulf has quit [Ping timeout: 248 seconds]
mkver has quit [Read error: Connection reset by peer]
mkver has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 248 seconds]
mkver has quit [Ping timeout: 244 seconds]
^Neo has quit [Ping timeout: 252 seconds]
^Neo_ has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
quietvoid has quit [Remote host closed the connection]
quietvoid has joined #ffmpeg-devel
lexano has quit [Remote host closed the connection]
lexano has joined #ffmpeg-devel
Anthony_ZO has quit [Ping timeout: 245 seconds]
jamrial has joined #ffmpeg-devel
ramiro has quit [Ping timeout: 252 seconds]
ramiro has joined #ffmpeg-devel
minimal has joined #ffmpeg-devel
<fflogger>
[editedticket] mfcc64: Ticket #11640 ([avfilter] Specific combination of timeclamp, fps, and count on showcqt causing crash) updated https://trac.ffmpeg.org/ticket/11640#comment:5
minimal has quit [Ping timeout: 244 seconds]
minimal has joined #ffmpeg-devel
IndecisiveTurtle has quit [Quit: IndecisiveTurtle]
MisterMinister has joined #ffmpeg-devel
TheVibeCoder has quit [Changing host]
TheVibeCoder has joined #ffmpeg-devel
System_Error has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 260 seconds]
MisterMinister has joined #ffmpeg-devel
<kasper93>
mkver: what's the status of your sanitizers related fixes? I recal you sent few patches some time ago. I want to clean fate warnings, but I think you fixes like 90% of it already.
<kasper93>
fixed*
<mkver>
I'll apply them.
<kasper93>
thanks
quietvoid has quit [Ping timeout: 244 seconds]
<TheVibeCoder>
JEEB: do you remmember any OOM bugs you reported on trac about filters?
MisterMinister has quit [Remote host closed the connection]
<TheVibeCoder>
found it
BradleyS has quit [Quit: quit]
BradleyS has joined #ffmpeg-devel
<kasper93>
TheVibeCoder: added this youtube cubemap format to my script, it was scaling issue, now it's perfect https://github.com/kasper93/mpv360
<TheVibeCoder>
kasper93: v360 have EAC already
<TheVibeCoder>
and myriad other transforms
<kasper93>
it's slow, shrug
IndecisiveTurtle has joined #ffmpeg-devel
<TheVibeCoder>
because its CPU
<TheVibeCoder>
write vulkan and/or opencl variant
<kasper93>
yeah
<TheVibeCoder>
I have half-working incomplete opencl version
cone-459 has joined #ffmpeg-devel
<cone-459>
ffmpeg Andreas Rheinhardt master:7574c5576138: avcodec/indeo3: Fix UB pointer arithmetic
<cone-459>
ffmpeg Andreas Rheinhardt master:41d86b5be5f3: avcodec/indeo3: Consistently use ptrdiff_t for strides
<cone-459>
ffmpeg Andreas Rheinhardt master:bdb4cd44e24c: avcodec/ffv1enc: Use dummies to avoid UB pointer arithmetic
<cone-459>
ffmpeg Andreas Rheinhardt master:134c1d81a43f: avcodec/indeo3: Constify ref_block in decode_cell_data()
<cone-459>
ffmpeg Andreas Rheinhardt master:2db60f90dfcd: swresample/rematrix: Constify
<cone-459>
ffmpeg Andreas Rheinhardt master:589f1e5cc728: swresample/rematrix: Use correct function pointer types for calls
chainik1 has quit [Quit: (╯°□°）╯︵ ┻━┻]
chainik15 has joined #ffmpeg-devel
<mkver>
jamrial: Do you know whether it is intended that the version strings (av_codec_ffversion etc.) are exported?
<jamrial>
i think that was added by michaelni, and that it was the intention
ngaullier has quit [Remote host closed the connection]
<kasper93>
TheVibeCoder: v360 is also hard to use for stereoscopic input, because you need to extract specific parts of image, crop, rotate and so on
<TheVibeCoder>
kasper93: in_stereo and out_stereo are options
<kasper93>
%z is supported since at least VS2015, not sure what is the minimum version to build ffmpeg
<kasper93>
though since we require C11 complier at least, I would guess older msvc would have other issues too
<cone-459>
ffmpeg Andreas Rheinhardt master:2845013154fa: tests/fate/screen: Add test for skipping cursor with FIC
<cone-459>
ffmpeg Andreas Rheinhardt master:5d5e922088fa: avcodec/fic: Avoid copying cursor unnecessarily
<cone-459>
ffmpeg Andreas Rheinhardt master:d6986e1fcdb2: avcodec/fic: Avoid implicit av_frame_free()+av_frame_alloc()
<cone-459>
ffmpeg Andreas Rheinhardt master:12ae987a6a7d: avcodec/fic: Don't copy frame unnecessarily
<cone-459>
ffmpeg Andreas Rheinhardt master:4c69f302a91b: avcodec/fic: Remove redundant logmessage
<cone-459>
ffmpeg Andreas Rheinhardt master:8bed944dda44: avcodec/fic: Don't prematurely claim to have decoded a frame
<cone-459>
ffmpeg Andreas Rheinhardt master:45920251d9e9: avcodec/fic: Ensure skip frames have up-to-date props
<cone-459>
ffmpeg Andreas Rheinhardt master:e813e1e940d3: avcodec/fic: Postpone duplicating AVFrame buffer
<mkver>
But removing SIZE_SPECIFIER also affects distributed binaries and not only systems used to build FFmpeg. Would this break some still supported systems?
<TheVibeCoder>
btw, great work!
<BtbN>
mkver: how would it affect distributed binaries? It's a macro?
<kasper93>
to run binaries you have to install vcredist (c/c++ stdlib) on target system compatible with your toolchain (or link statically), so compatiblity is carried over.
<BtbN>
Support for printing it at runtime comes from the MSVCRT, and naturally stuff build with VS2015 will need the MSVCRT140
<BtbN>
which supports it
<kasper93>
as for suported systems, nothing would be affected, becase even windows10 will soon not be supported.
<BtbN>
Win10 got at least 3 more years in it
<BtbN>
They made the extended support effectively free now
<kasper93>
though people on doom9 are still sharing binaries for WinXP ... so there is that
<mkver>
kasper93: Really? I thought we still support everything Vista+.
<BtbN>
I know that Win7 does not work anymore, judging by the constant complaints from people on my builds, despite the README clearly saying I only care about supported Windows Versions
<BtbN>
Though it's near impossible to tell if that's our fault or some other dependency being linked in
<kasper93>
it should yell, "procedure X is not found" if that's the issue
<BtbN>
Usually the best info I get is "it doesn't work"
<kasper93>
yeah, what do you expect
<BtbN>
and I don't feel like picking for info for something I have no intention to fix
<TheVibeCoder>
ffmpeg doesn't usually work
<kasper93>
we got very nasty messages on mpv issues when windows xp and later 7 stopped working
<TheVibeCoder>
btw, fps filter in ffmpeg is balloning memory all the time
<kasper93>
looks like configure sets _WIN32_WINNT=0x0600 so Vista+ is "supported", in theory.
<BtbN>
My builds set the version of the mingw headers and libraries themselves to Win7+
<BtbN>
But clearly something is unhappy on actual Win7
<BtbN>
And I honestly couldn't care less about it
<jamrial>
kasper93: libavcodec/d3d11va.h for example forces it to be 0x602 (Win8)
<jamrial>
also dxva2.h
<jamrial>
so does configure, so i guess that's why BtbN's builds wont work on win7
<BtbN>
so lets just bump configure to Win8.1 to be clear about it?
<TheVibeCoder>
what is latest windows os version? 12 ?
<jamrial>
if you don't enable dxva2, d3d11va, mfenc and vsrc_ddagrab, it will work on Vista+ in theory
<jamrial>
TheVibeCoder: 11
<cone-459>
ffmpeg Andreas Rheinhardt master:baccb96fce6b: avcodec/ituh263dec: Use correct logcontext
<cone-459>
ffmpeg Andreas Rheinhardt master:ba79fff6d523: avcodec/rl: Avoid branch in index lookup
<cone-459>
ffmpeg Andreas Rheinhardt master:8a3566498e57: avcodec/ituh263enc: Simplify creating LUT
<cone-459>
ffmpeg Andreas Rheinhardt master:1c258b558120: avcodec/ituh263dec: Only initialize ff_h263_rl_inter when needed
<cone-459>
ffmpeg Andreas Rheinhardt master:3b87a480fe88: avcodec/mpegvideoenc: Allocate blocks as part of MPVEncContext
<cone-459>
ffmpeg Andreas Rheinhardt master:0723fcce51f8: avcodec/mpegvideo: Add MPVContext typedef
<cone-459>
ffmpeg Andreas Rheinhardt master:261d8eb5ddf0: avcodec/mpegvideo_dec: Factor debugging dct coefficients out
<cone-459>
ffmpeg Andreas Rheinhardt master:8cff637cf0de: avcodec/mpegvideo_dec: Reindent after the previous commit
<cone-459>
ffmpeg Andreas Rheinhardt master:4fc874ef0813: avcodec/mpeg_er: Don't use MpegEncContext.block
<cone-459>
ffmpeg Andreas Rheinhardt master:fabbe7917078: avcodec/mpegvideodec: Remove size expectation from ff_mpv_reconstruct_mb
<cone-459>
ffmpeg Andreas Rheinhardt master:80ed8682967e: avcodec/h261dec: Stop using MpegEncContext.gb
<cone-459>
ffmpeg Andreas Rheinhardt master:21c88b29f7ff: avcodec/h261dec: Don't use MpegEncContext.block
<cone-459>
ffmpeg Andreas Rheinhardt master:b38f23663e76: avcodec/mpeg12dec: Put GetBitContext on the stack where advantageous
<cone-459>
ffmpeg Andreas Rheinhardt master:86ee71f4e989: avcodec/mpeg12dec: Remove unused function parameter
<cone-459>
ffmpeg Andreas Rheinhardt master:0e6d59871f17: avcodec/rv34: Don't use MpegEncContext.gb
<cone-459>
ffmpeg Andreas Rheinhardt master:25e4baeed488: avcodec/rv34: Don't use MpegEncContext.block
<cone-459>
ffmpeg Andreas Rheinhardt master:aa4a5b1e3306: avcodec/intrax8: Don't pretend to need more than one int16_t[64]
<cone-459>
ffmpeg Andreas Rheinhardt master:fe2dbb12b07e: avcodec/vc1: Stop using MpegEncContext.gb
<cone-459>
ffmpeg Andreas Rheinhardt master:94cfbf02385e: avcodec/vc1: Don't use MpegEncContext.block
<cone-459>
ffmpeg Andreas Rheinhardt master:ee81e284693d: avcodec/mpeg12dec: Deduplicate variables
<cone-459>
ffmpeg Andreas Rheinhardt master:05510dac6b84: avcodec/mpegvideo: Move flipflop_rounding to {MSMPEG4Dec,MPVEnc}Context
<cone-459>
ffmpeg Andreas Rheinhardt master:273de7f60c04: avcodec/mpegvideo: Move unrestricted_mv to MotionEstContext
<cone-459>
ffmpeg Andreas Rheinhardt master:91e9578025f8: avcodec/mpeg4videodec: Avoid unnecessary indirections
<cone-459>
ffmpeg Andreas Rheinhardt master:758c8dffc922: avcodec/{h263,mpeg4video}dec: Pass MPVContext*, not Mpeg4DecContext*
<cone-459>
ffmpeg Andreas Rheinhardt master:3f5c2043cc41: avcodec/mpegvideo: Move dct_precision to Mpeg4DecContext
<cone-459>
ffmpeg Andreas Rheinhardt master:584d50c5f3df: avcodec/h263dec: Add H263DecContext
<cone-459>
ffmpeg Andreas Rheinhardt master:b2be342ab668: avcodec/h263dec: Remove redundant block parameter from decode_mb
<cone-459>
ffmpeg Andreas Rheinhardt master:77add26918fb: avcodec/h263dec: Don't use MpegEncContext.block
<cone-459>
ffmpeg Andreas Rheinhardt master:8a4accc24dca: avcodec/h263dec: Stop using MpegEncContext.gb
<cone-459>
ffmpeg Andreas Rheinhardt master:3a96a3c949a9: avcodec/mpeg12dec: Add Mpeg12SliceContext
<cone-459>
ffmpeg Andreas Rheinhardt master:27d785746c37: avcodec/mpegvideo: Add missing headers
<cone-459>
ffmpeg Andreas Rheinhardt master:e40b0ebb2e5d: avcodec/mpeg12dec: Move MpegEncContext.gb to Mpeg12SliceContext
<cone-459>
ffmpeg Andreas Rheinhardt master:f1c56c08ed3b: avcodec/mpeg12dec: Don't use MPVContext.block
<cone-459>
ffmpeg Andreas Rheinhardt master:7db6267b30b2: avcodec/mpegvideo: Move fields only used by H.263 decoders to H263DecCtx
<cone-459>
ffmpeg Andreas Rheinhardt master:9f42725d7526: avcodec/mpegvideo: Move mb_num_left to {H263,RV34}DecContext
<cone-459>
ffmpeg Andreas Rheinhardt master:d5fa28293606: avcodec/mpeg12dec: Put mb_skip_run on the stack
<cone-459>
ffmpeg Andreas Rheinhardt master:2089d39735c1: avcodec/mpegvideo: Move mb_skip_run to {RV34Dec,MPVEnc}Context
<cone-459>
ffmpeg Andreas Rheinhardt master:502a6ea123cf: avcodec/mpegvideo: Move SLICE_* defs to h263dec.h, h261dec.c
<cone-459>
ffmpeg Andreas Rheinhardt master:3b6d1a6ee129: avcodec/msmpeg4dec: Move ff_msmpeg4_decode_init() down
<cone-459>
ffmpeg Andreas Rheinhardt master:c660f45f8e6a: avcodec/h263dec: Use function ptr for decode_picture_header
<cone-459>
ffmpeg Andreas Rheinhardt master:532ae5dc7da4: avcodec/ituh263enc: Inline value of h263_flv
<cone-459>
ffmpeg Andreas Rheinhardt master:00ea78f4264a: avcodec/flvdec: Binarize h263_flv
<cone-459>
ffmpeg Andreas Rheinhardt master:3c06d87c93e0: avcodec/mpegvideo: Move fields to {H263Dec,MPVEnc}Context when possible
<cone-459>
ffmpeg Andreas Rheinhardt master:0984724d1e6d: avcodec/mpeg_er: Allow to skip setting partitioned_frame, p[pb]_time
<cone-459>
ffmpeg Andreas Rheinhardt master:60f51bdaac29: avcodec/mpegvideo: Move partitioned_frame to {H263Dec,MPVEnc}Context
<cone-459>
ffmpeg Andreas Rheinhardt master:abc54c25befa: avcodec/mpegvideo: Move loop_filter to {H263Dec,MPVEnc,VC1}Context
<cone-459>
ffmpeg Andreas Rheinhardt master:fa02dea036dd: avcodec/rv34: Don't report progress unnecessarily
<cone-459>
ffmpeg Andreas Rheinhardt master:01ffe103e502: avcodec/rv34: Fix spelling mistake
<TheVibeCoder>
big beautiful committ
Everything has joined #ffmpeg-devel
<BtbN>
Wasn't av_log(NULL, ...) invalid?
<jamrial>
no
<BtbN>
I vagule remember that it could lead to crashes with custom log handlers
<mkver>
It's valid, but discouraged.
<mkver>
No. Using av_log(ctx, ...) with *(void**)ctx == NULL could lead to crashes.
<mkver>
What makes you bring up av_log(NULL?
<kasper93>
BtbN: everyone jumped over win8, browsers and other big guys. Reason being, while win7 was still big target, win8 is basically extinct so 0x0A00 is probably next logical step
<BtbN>
mkver: the WHIP code uses it everywhere, whenever it outsourced code elsewhere
<BtbN>
see all the stuff it put into tls_openssl
<BtbN>
And it also in turn forced me to do the same, to follow the internal API it designed, since the functions it wants me to implement get no log context, but absolutely want to log
bwu25 has quit [Quit: bwu25]
<wbs>
jamrial: regarding d3d11 and stuff manually defining newer versions of windows - I think that's with code that explicitly dynamically loads any symbol, so it shouldn't require that version at runtime; that's the intent at least
<jamrial>
wbs: mm, then no idea why a build wouldn't work on win7
<jamrial>
maybe some external dependency is at fault
<wbs>
jamrial: yeah, that's probably most plausible. I could try, I actually had a win7 vm up the other day
<wbs>
regarding %z, msvcrt.dll printf didn't support it back in the day. but ffmpeg's configure explicitly requests the mingw-provided custom printf reimplementations, which does support it
<jamrial>
and i'm fairly sure we don't support old msvc at this point
<jamrial>
like, 2015 as the oldest or so
<wbs>
yes, for other reasons than that, we require msvc 2019 or so, I think
user23 has joined #ffmpeg-devel
user23 has quit [Excess Flood]
user23 has joined #ffmpeg-devel
<BtbN>
no need for the clunky printf helper then
Coinflipper has quit [Quit: ]
Coinflipper has joined #ffmpeg-devel
user23 has left #ffmpeg-devel [#ffmpeg-devel]
<wbs>
yeah we could probably get rid of that these days
odrling has quit [Remote host closed the connection]
odrling has joined #ffmpeg-devel
<cone-459>
ffmpeg Marton Balint master:f3b7aa6c224d: avfilter: factorize requesting an input frame from multi output filters
<averne>
Like, if it was just a morton curve you could say that's because it's simple to compute so you don't need a LUT. But I can't make sense of this
<averne>
Patents maybe?
Everything has quit [Quit: leaving]
<beastd>
TheVibeCoder: Not exactyl sure but I would say this intuitively also do it like this: #define FF_PTR_ADD(ptr, off) (!(ptr) ? (ptr) : (ptr) + (off))
<kasper93>
mkver: Could you also apply stack-buffer-overflow fix for checkasm/h264dsp? It has LGTM on ML.
<beastd>
First usage was of FF_PTR_ADD was like this, so it's working fine if you assume that the offset for unset frame planes evaluates to zero: out[i] = FF_PTR_ADD(out_buf->data[i], field * out_buf->linesize[i]);
<beastd>
michaelni,mkver: Do you remember the reasoning behind checking off and not ptr? (avutil/common: Add FF_PTR_ADD() Suggested-by: Andreas Rheinhardt)
<BtbN>
mkver, wbs are we fine with pushing %zu? Or should I make the patch use the macro anyway, for consistency?
<wbs>
BtbN: I don't have a strong opinion on it. for consistency it could of course be nice to keep doing what we're doing elsewhere, and then take a separate step to stop using that, with clarifications about why we no longer need to do that. but no strong opinion from me
<mkver>
beastd: It allows to optimize the branch away. GCC does it, Clang does not (last I checked).
<mkver>
BrbN: I don't care.
<BtbN>
I'll push it like this then, less noise, and if we ever remove it, it can be explained either way
iive has joined #ffmpeg-devel
<cone-459>
ffmpeg Timo Rothenpieler master:ba984355fe5c: avformat/whip: fix format string for printing size_t
<cone-459>
ffmpeg Lidong Yan master:7ed5a7094fe6: avformat/rtpdec_latm: fix leak in parse_fmtp_config()
<cone-459>
ffmpeg Lidong Yan master:98afcd3da776: avcodec/sunrast: fix leak in sunrast_decode_frame()
<cone-459>
ffmpeg Lidong Yan master:e68599f551de: avformat/rtpdec_asf: fix leak in ff_wms_parse_sdp_a_line()
<cone-459>
ffmpeg Michael Niedermayer master:ecbe3e73662b: avcodec/rv60dec: Check ofs for overflows
quietvoid has joined #ffmpeg-devel
<beastd>
mkver: thanks. understood. tried a bit with newish clang and gcc for aarch64 and x86-64 and indeed the generated code is favourable for the current variant of FF_PTR_ADD
<beastd>
makes the macro less broad than it name seems to suggest, but for the supported cases it seems better on popular compilers and architectures (well not for MSVC)
<kasper93>
this macro is useful for NULL+0 cases, which incidentally in longer reported by newer clang. NULL+N are more interesting
<beastd>
kasper93: yeah, i figured similar. tho actually not sure if NULL+0 is really allowed by C std. tho it seems rather silly to disallow addition of zero even for NULL because it shoudn't change anything anyway
<mkver>
beastd: NULL + 0 is definitely UB before C23. I don't know about C23 though.
<mkver>
This means FF_PTR_ADD is unnecessary for C23.
<beastd>
so actually at some point in the future we could do the following dance (partially): 1) we could remove FF_PTR_ADD at all places where it's currently used and ḿaybe 2) define a broader FF_PTR_ADD version to avoid NULL pointer arith when we see a need for it
wbs has joined #ffmpeg-devel
<mkver>
If by "broader" you mean something that checks against NULL + something !=0, then I don't think that we need a macro for this.
<beastd>
yes, that's what thought. and also think it's probably not needed. at least can't remember any instances where it would be handy.
<kasper93>
FF_PTR_ADD could be defined as "checked" add for <C23 and as simple addition for C23>=0, if we want to be very strict about this.
<beastd>
possible, but actually I think just getting rid of it at some point (whenever that will be the case) is fine. because the current definition of FF_PTR_ADD is good enough for gcc and clang to emit optimal code for popular targets
<kasper93>
graphparser.c:353 is leaking, unfortunatelly oss-fuzz is useless for oom, but it crashed with 13230880147 byte(s) (99%) in 61186 allocation(s) allocated from this
<TheVibeCoder>
cmd line?
System_Error has quit [Ping timeout: 244 seconds]
<kasper93>
it's some dumb fuzzed string, let me try to repro locally