[Ffmpeg-devel-irc] ffmpeg-devel.log.20180618

burek burek021 at gmail.com
Tue Jun 19 03:05:03 EEST 2018

[03:04:06 CEST] <cone-730> ffmpeg 03Xiaohan Wang 07release/3.4:988e7a8feaf3: avcodec/h264_cavlc: Set valid qscale value in ff_h264_decode_mb_cavlc()
[03:04:06 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:a1e983111ee3: avcodec/h264_cabac: Tighten allowed coeff_abs range
[03:04:06 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:1a8265487b89: avdevice/decklink_dec: Fix ;;
[03:04:06 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:3c001b18cea7: avutil/common: Fix integer overflow in av_clip_uint8_c() and av_clip_uint16_c()
[03:04:06 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:fe435d11fb59: avcodec/exr: Check remaining bits in last get code loop
[03:04:06 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:3f6f1c24ebee: avcodec/vp8: Check for bitstream end before vp7_fade_frame()
[03:04:06 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:fcc6932e0984: avcodec/jpeg2000dec: Use av_image_check_size2()
[03:04:07 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:4597b51e8a43: avcodec/hevcdec: Check luma/chroma_log2_weight_denom
[03:04:08 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:ad3ef9e89609: avcodec/dirac_dwt_template: Fix Integer overflow in horizontal_compose_dd137i()
[03:04:09 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:3fc7e88bfd8f: avcodec/dxtory: Remove code that corrupts dimensions
[03:04:10 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:5c2e61f1f527: avcodec/diracdec: Use int64 in global mv to prevent overflow
[03:04:11 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:c8f54bf87cde: avcodec/jpeg2000dwt: Fix integer overflows in sr_1d53()
[03:04:12 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:28398a70a975: avcodec/aacdec_templat: Fix integer overflow in apply_ltp()
[03:04:13 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:bc194927e4ae: avcodec/h264_parse: Clear invalid chroma weights in ff_h264_pred_weight_table()
[03:04:14 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:2b5c0fdaaca5: avcodec/diracdec: Fix integer overflow in mv computation
[03:04:15 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:411310659591: avcodec/cavsdec: Check alpha/beta offset
[03:04:16 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:30a718c38b62: avcodec/smc: Check input packet size
[03:04:17 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:ae49cc73f265: avcodec/utvideodec: Check subsample factors
[03:04:18 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:ee33e79b25b8: avcodec/bintext: sanity check dimensions
[03:04:19 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:60bd8316db46: avcodec/exr: fix invalid shift in unpack_14()
[03:04:20 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:9cc83b94a1f8: avcodec/g2meet: Check tile dimensions with av_image_check_size2()
[03:04:21 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:8456bba36297: avcodec/truemotion2rt: Check input buffer size
[03:04:22 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:7a713cea22f4: avformat/mov: Fix integer overflow in mov_get_stsc_samples()
[03:04:23 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:a05d5770682d: libavformat/oggparsevorbis: Fix memleak on multiple headers
[03:04:24 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:cd7a2954ae74: avformat/oggdec: Fix metadata memleak on multiple headers
[03:04:25 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:fea47f6941a5: avformat/utils: Fix integer overflow of fps_first/last_dts
[03:04:26 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:60f676712600: avformat/oggparsetheora: Do not adjust AV_NOPTS_VALUE
[03:04:27 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:79efbd547ea3: avformat/avidec: Fix integer overflow in cum_len check
[03:04:28 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:a376fcaa39d0: avformat/oggparseogm: Fix undefined shift in ogm_packet()
[03:04:29 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:0cdc3e05cbc8: avformat/oggparseogm: Check lb against psize
[03:04:30 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:24c627a90099: avformat/oggparsedaala: Do not adjust AV_NOPTS_VALUE
[03:04:31 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:3fa6e594a0f2: avformat/img2dec: fix infinite loop
[03:04:32 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:828044878484: avformat/mov: Fix integer overflows related to sample_duration
[03:04:33 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:8ba8e0a1c773: avcodec/wmalosslessdec: Reset num_saved_bits on error path
[03:04:34 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:031206032960: avcodec/nuv: Check for minimum input size for uncomprssed and rtjpeg
[03:04:35 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:46ed4d528253: avcodec/nuv: rtjpeg with dimensions less than 16 would result in no decoded pixels thus reject it
[03:04:36 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:e26ae6e6cec8: avformat/mov: Check STSC and remove invalid entries
[03:04:37 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:26fcff1eb40d: avcodec/get_bits: Make sure the input bitstream with padding can be addressed
[03:04:38 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:9b2ce61155db: avformat/mov: Move +1 in check to avoid hypothetical overflow in add_ctts_entry()
[03:04:39 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:2be397e234a0: avcodec/tableprint_vlc: Fix build failure with --enable-hardcoded-tables
[03:04:40 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:fdf56c5654c8: avcodec/wmalosslessdec: Fix null pointer dereference in decode_frame()
[03:04:41 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:51035698bde9: avcodec/utvideodec: Set pro flag based on fourcc
[03:04:42 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:43916494f8ca: avfilter/vf_signature: use av_strlcpy()
[03:04:43 CEST] <cone-730> ffmpeg 03heimdallr 07release/3.4:2c2ab8556263: avcodec/imgconvert: Fix loss mask bug in avcodec_find_best_pix_fmt_of_list()
[03:04:44 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:f4e01b87f2bc: avcodec/cscd: Error out when LZ* decompression fails
[03:04:45 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:46db1182da23: avcodec/aacdec_fixed: Fix integer overflow in apply_independent_coupling_fixed()
[03:04:46 CEST] <cone-730> ffmpeg 03Hendrik Schreiber 07release/3.4:3e6adf0842b6: swresample/swresample: Fix for seg fault in swr_convert_internal() -> sum2_float during dithering.
[03:04:47 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:c8f700975ea0: libavcodec/rv34: error out earlier on missing references
[03:04:48 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:053b9766b4a9: avcodec/aacsbr_fixed: Fix integer overflow in sbr_hf_assemble()
[03:04:49 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:c44244424ff2: avcodec/movtextdec: Check style_start/end
[03:04:50 CEST] <cone-730> ffmpeg 03Matt Wolenetz 07release/3.4:ae136f8d6425: lavc/libopusdec: Allow avcodec_open2 to call .close
[03:04:51 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:0f868badcfdb: avformat/mov: Fix extradata memleak
[03:04:52 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:fffc77b47280: avcodec/h264_slice: Fix integer overflow with last_poc
[03:04:53 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:78c8e77c8635: avcodec/mjpegdec: Check input buffer size.
[03:04:54 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:ca119e9456a2: avformat/utils: Fix integer overflow in end time calculation in update_stream_timings()
[03:04:55 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:8d05a802dd2f: avcodec/truemotion2: Propagate out of bounds error from GET_TOK()
[03:04:56 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:22a4b599f05e: avcodec/utils: Enforce minimum width also for VP5/6
[03:04:57 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:0b5a2c9a52eb: avformat/utils: Check cur_dts in update_initial_timestamps() more
[03:04:58 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:86a6e700e116: doc/APIchanges: Fix typos in hashes
[03:04:59 CEST] <cone-730> ffmpeg 03Rahul Chaudhry 07release/3.4:47c3d2459e95: swresample/arm: remove unintentional relocation.
[03:05:00 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:205e93fa0d2b: avcodec/cinepak: move some checks prior to frame allocation
[03:05:01 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:a6db663a31ce: avcodec/cinepak: Skip empty frames
[03:05:02 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:86c7035646d5: avcodec/dfa: Check dimension against maximum
[03:05:03 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:81d8bd6f421b: avcodec/dsicinvideo: Propagate errors from cin_decode_rle()
[03:05:04 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:f93d7a0c0a23: avcodec/dsicinvideo: Fail if there is only a small fraction of the data available that comprises a full frame
[03:05:05 CEST] <cone-730> ffmpeg 03Stephan Holljes 07release/3.4:1d5694ba1101: lavf/http.c: Free allocated client URLContext in case of error.
[03:05:06 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:be9c4cc0e60a: avcodec/ffv1enc: Check that the crc + version combination is supported
[03:05:07 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:f31ead3f76da: avcodec/vc1_block: simplify ac_val computation
[03:05:08 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:24947b3f7dd2: avcodec/elsdec: Fix memleaks
[03:05:09 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:3b23dd4b91af: avcodec/h263dec: Check slice_ret in mspeg4 slice loop
[03:05:10 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:30ec2357c98a: avcodec/error_resilience: Fix integer overflow in filter181()
[03:05:11 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:c437510d5cf6: avcodec/wavpack: Fix integer overflow in wv_unpack_stereo()
[03:05:12 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:909c1aa0ec27: avcodec/wavpack: Fix integer overflow in DEC_MED() / INC_MED()
[03:05:13 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:5d1c783bf027: avcodec/flac_parser: Fix infinite loop
[03:05:14 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:2cb3dff0f938: avcodec/g2meet: Change order of operations to avoid undefined behavior
[03:05:15 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:f4d90a36c5d3: avcodec/jpeg2000dec: Skip init for component in CPRL if nothing is to be done
[03:05:16 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:022c8c07f82d: avcodec/jpeg2000dec: Fix undefined shift in the jpeg2000_decode_packets_po_iteration() CPRL case
[03:05:17 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:b802baadfda4: avcodec/g2meet: Check RGB upper limit
[03:05:18 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:14cea440183f: avcodec/fic: Check available input space for cursor
[03:05:19 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:22b3e924dfd6: avcodec/dirac_dwt: Fix integer overflow in COMPOSE_DD97iH0 / COMPOSE_DD137iL0
[03:05:20 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:3712d4da1e2c: avformat/mov: Only fail for STCO/STSC contradictions if both exist
[03:05:21 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:9d3ae9051622: indeo4: Decode all or nothing of a band header.
[03:05:22 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:4cbd92baa3f4: avcodec/mlpdec: Only change noise_type if the related fields are valid
[03:05:23 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:bd1fd3ff4b04: avcodec/mpeg4videoenc: Use 64 bit for times in mpeg4_encode_gop_header()
[03:05:24 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:c91f199f5d19: avcodec/g723_1dec: Clip bits2 in both directions
[03:05:25 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:b7b178fd0fa1: oavcodec/aacpsdsp_template: Use unsigned for hs0X to prevent undefined behavior
[03:05:26 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:ada22c1c0f55: avcodec/aacdec_fixed: use 64bit to avoid overflow in rounding in apply_dependent_coupling_fixed()
[03:05:27 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:12f0cfede607: avcodec/g2meet: ask for sample with overflowing RGB
[03:05:28 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:26667dc5baaf: avcodec/fic: Avoid some magic numbers related to cursors
[03:05:29 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:ce7be9cdc5d5: avformat/mov: Break out early if chunk_count is 0 in mov_build_index()
[03:05:30 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:adea365be9e0: avformat/mov: replace a value error by clipping into valid range in mov_read_stsc()
[03:05:31 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:f0c60a91ae86: avcodec/amrwbdec: Fix division by 0 in find_hb_gain()
[03:05:32 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:df7d70272e90: avcodec/opus_silk: Change silk_lsf2lpc() slightly toward silk/NLSF2A.c
[03:05:33 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:c4c56eb839c0: avcodec/truemotion2: Fix overflow in tm2_apply_deltas()
[03:05:34 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:aa30aa8c1364: avcodec/mjpegdec: Fix integer overflow in ljpeg_decode_rgb_scan()
[03:05:35 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:7598b161f4ee: avcodec/shorten: Check non COMM chunk len before skip in decode_aiff_header()
[03:05:36 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:3f5bc91be324: avcodec/shorten: Sanity check nmeans
[03:05:37 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:8b6c773fe80d: avcodec/shorten: Fix a negative left shift in shorten_decode_frame()
[03:05:38 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:62e1c5db4acc: avcodec/shorten: Fix undefined shift in fix_bitshift()
[03:05:39 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:5d6140074084: avcodec/shorten: Fix multiple integer overflows
[03:05:40 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:b60a750af758: avcodec/wavpack: Fix overflow in adding tail
[03:05:41 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:5d6d0d20fe3f: avcodec/xwddec: Use ff_set_dimensions()
[03:05:42 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:28f044965975: avcodec/h264_mc_template: Only prefetch motion if the list is used.
[03:05:43 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:3be0549029b0: avcodec/h264_ps: Move MAX_LOG2_MAX_FRAME_NUM to header so it can be used in h264_sei
[03:05:44 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:78fe6f7c32dd: avcodec/h264_slice: Fix overflow in recovery_frame computation
[03:05:45 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:532ba2e4ae8c: avformat/mov: Only set pkt->duration to non negative values
[03:05:46 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:f35fb0c82692: fftools/ffmpeg: Fallback to duration if sample rate is unavailable
[03:05:47 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:84d473a9cfc6: avutil/common: Fix undefined behavior in av_clip_uintp2_c()
[03:05:48 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:d946e7b805e2: avcodec/dirac_dwt_template: Fix undefined behavior in interleave()
[03:05:49 CEST] <cone-730> ffmpeg 03Michael Niedermayer 07release/3.4:41315d5c7d48: avcodec/aacdec_fixed: Fix undefined integer overflow in apply_independent_coupling_fixed()
[03:27:19 CEST] <jamrial_> fate is awfully yellow
[03:27:28 CEST] <jamrial_> something broke fate-mss2-wmv
[03:28:54 CEST] <jamrial_> oh, michaelni already reported it
[11:59:33 CEST] <cone-223> ffmpeg 03Carl Eugen Hoyos 07master:a707a0b657da: Revert "avcodec/vc1: add Simple and Main profile to vc1_put_signed_blocks_clamped"
[11:59:34 CEST] <cone-223> ffmpeg 03Jerome Borsboom 07master:4dc7ff7e104b: avcodec/vc1: add Simple and Main profile to vc1_put_signed_blocks_clamped
[11:59:35 CEST] <cone-223> ffmpeg 03Jerome Borsboom 07master:c68f60e80a19: avcodec/vc1: rewrite vc1_decode_i_blocks to align with VC-1 spec
[11:59:36 CEST] <cone-223> ffmpeg 03Jerome Borsboom 07master:e85c608a4644: avcodec/vc1: remove unused ff_vc1_loop_filter_iblk
[13:01:47 CEST] <akravchenko188> jkqxz: Hi, I have sent two patches with amf hwcontext and vf_scale_amf. Could you please review them? Thanks.
[14:08:00 CEST] <thardin> reading the build system thread, why is configure so damn slow?
[14:08:09 CEST] <thardin> is it just that it is stateless?
[14:08:44 CEST] <thardin> I presume ffconfigure doesn't do all the useless checks for obsolete systems that autoconf's does
[14:14:33 CEST] <nevcairiel> its slow because shell is terrible for what it does
[14:14:54 CEST] <thardin> I doubt bash is the actual bottleneck
[14:15:06 CEST] <nevcairiel> the forking of a billion sub-process is the bottleneck
[14:15:14 CEST] <nevcairiel> instead of just having built-in functionality
[14:16:41 CEST] <nevcairiel> and thats not about testing compilers etc, obviously you have to use a separate process for that
[14:16:45 CEST] <nevcairiel> but basic string processing etc
[14:17:12 CEST] <thardin> hm, right
[14:21:32 CEST] <thardin> running some basic profiling on it, let's see what it spits out
[14:43:15 CEST] <thardin> hrm my wm died
[14:46:54 CEST] <thardin> OOM killer
[14:49:57 CEST] <thardin> death by a thousand forks
[14:53:50 CEST] <thardin> and no output to see for it
[14:58:08 CEST] <thardin> but lots of string concatenation which is likely O(N^2)
[15:59:15 CEST] <thardin> 7 minutes wall time
[15:59:24 CEST] <thardin> ridic
[16:07:19 CEST] Action: Mathieu_Du interested about configure profiling too
[16:08:26 CEST] <thardin> I went by the second suggestion here: https://stackoverflow.com/questions/5014823/how-to-profile-a-bash-shell-script-slow-startup
[16:08:34 CEST] <thardin> the resulting dump is pretty thicc
[16:08:54 CEST] <thardin> diffing, sorting and cutting lines down to 80 chars -> 205M
[16:09:09 CEST] <thardin> and it's of moderate use I think
[16:09:46 CEST] <thardin> top five:
[16:10:01 CEST] <thardin>  0.409855126    + enable makeinfo
[16:10:01 CEST] <thardin>  0.363928815    ++ echo
[16:10:01 CEST] <thardin>  0.311682872    + append CXXFLAGS -std=c++11
[16:10:01 CEST] <thardin>  0.269393447    + '[' 06 -ge 5 ']'
[16:10:01 CEST] <thardin>  0.138258064    + enable pod2man
[16:11:28 CEST] <thardin> enable() and append() are both quite trivial in what they do
[16:12:33 CEST] <thardin> if it were at least bash then lists and hashes could be used. but it's sh
[16:15:35 CEST] <thardin> I'll post my results on the ML. a middle step might be to make use of said bashisms. will probably irate some
[16:17:56 CEST] <Mathieu_Du> thardin, or leave optimization concerns to a third party ^^
[16:18:56 CEST] <thardin> well this is kind of annoying when working on say new codecs
[16:19:14 CEST] <thardin> having a *seven minute* interruption to reconfigure is simply not acceptable
[16:19:29 CEST] <Mathieu_Du> yes
[16:19:37 CEST] <Mathieu_Du> did you try the meson port ooc ?
[16:19:49 CEST] <thardin> I did not
[16:19:56 CEST] <thardin> I'm just kind of laying out how things are now
[16:20:08 CEST] <Mathieu_Du> right
[16:20:55 CEST] <Mathieu_Du> fwiw meson build time is similar for me on all three platforms, at around 15 seconds
[16:21:18 CEST] <thardin> noice
[16:21:58 CEST] <thardin> I presume this is just the configure step
[16:25:03 CEST] <Mathieu_Du> Yes of course, compile time is pretty much the same
[16:27:01 CEST] <Mathieu_Du> ah I see how that might have been confusing, I meant the time taken by running "meson build"
[16:27:02 CEST] <thardin> come to think of it, how fast does a "make" with zero changes take?
[16:27:20 CEST] <thardin> tup and ninja are pretty snappy when it comes to that
[16:27:33 CEST] <Mathieu_Du> that would be ninja -C build, it's pretty much instant
[16:29:10 CEST] <thardin> mm
[16:29:17 CEST] <thardin> I use cmake+ninja at work
[16:29:48 CEST] <thardin> cmake+make is slightly slow, but at least it isn't msbuild
[16:31:29 CEST] <nevcairiel> at least make runs fully parallel for ffmpeg, so there is nothing really wrong with the makefiles, just configure
[16:32:39 CEST] <thardin> ye. cmake spits out multiple makefiles unfortunately. that will always be slow I suspect
[16:32:59 CEST] <thardin> cmake's syntax is godawful sadly
[16:36:30 CEST] <thardin> mail sent
[16:39:33 CEST] <thardin> herp forgot to mention one needs to change configure to use bash. but still
[16:39:44 CEST] <Mathieu_Du> thardin, note that configure time is a nice side effect of the meson port, but IMO the main benefit really would be maintainability and readability
[16:41:04 CEST] <thardin> that I don't doubt
[16:41:51 CEST] <January> thardin: proposing adopting meson was more as a whole thing than only looking at configure performance
[16:42:06 CEST] <January> (though configure performance is a bit part of it)
[16:42:26 CEST] <thardin> meson.build is certainly much smaller
[16:42:56 CEST] <thardin> does meson do stuff like print every missing dep on the first go?
[16:43:00 CEST] <thardin> that's something cmake is missing
[16:43:15 CEST] <thardin> and autoconf
[16:43:25 CEST] <j-b> meson > cmake
[16:43:55 CEST] <j-b> meson is what Cmake should have been
[16:43:58 CEST] <thardin> I suspect cmake only hangs on because of msvc compat
[16:44:16 CEST] <thardin> literally anything would be better. scons maybe
[16:46:45 CEST] <Mathieu_Du> meson works with msvc fwiw
[16:49:05 CEST] <Mathieu_Du> re every missing dep, no, that's probably a bit tricky to do because depending on whether a dep is present or not, different conditions might be entered
[16:49:34 CEST] <thardin> right
[16:50:25 CEST] <Mathieu_Du> eg https://pastebin.com/fwLZPVyG
[18:49:01 CEST] <kurosu> Is meson 0.46 sufficient for the proposed build system? Because that's what is available under msys2
[18:50:31 CEST] <__tim> kurosu, unfortunately you need meson from git master for the time being (but a new release 0.47 should be out soon hopefully)
[18:50:43 CEST] <kurosu> ok
[18:50:47 CEST] <kurosu> otherwise, right, configure has become a bother
[18:51:02 CEST] <January> kurosu: meson is fairly easy to install from master
[18:51:04 CEST] <__tim> you should be able to run Meson irectly from a git checkout though ~/Devel/meson/meson.py builddir
[18:51:14 CEST] <January> or that
[18:51:51 CEST] <kurosu> but I suspect a lot of ffmpeg devs to be minimalists (see eg the pkgconfig eternal debate)
[18:52:10 CEST] <kurosu> I'd wager there could be some opposition
[18:52:24 CEST] <kurosu> as for me, I hardly ever run configure nowadays
[18:53:17 CEST] <January> kurosu: minimalists with a 10000 line custom configure?
[18:53:26 CEST] <kurosu> as in system dependencies
[18:54:04 CEST] <Shibe> could anybody help me? I can't seem to be able to get device constraints https://gist.github.com/123e67b34b94bf0a88508abb07fd91fc
[18:54:11 CEST] <Shibe> i dont know what im doing wrong
[18:54:12 CEST] <kurosu> there have been jokes about ffcc for that reason
[19:08:46 CEST] <jkqxz> Shibe:  The DRM hwcontext doesn't support internal frame allocation.  Usually it's used for mapping; if you want direct allocation then you need to implement it yourself.
[19:54:46 CEST] <Shibe> jkqxz: seems like switching hardware device type from drm to vaapi works
[19:54:55 CEST] <Shibe> thanks 
[20:16:02 CEST] <Mithgol> Could anybody help me? I'd like the regression https://twitter.com/FidonetRunes/status/1008451439893581824 that I observed using binaries https://twitter.com/FidonetRunes/status/1008452816049000450 to be filed as an appropriate ticket on FFmpeg's trac.
[20:38:39 CEST] <kurosu> 4m15 vs 45s on Win7 x64 with no option
[20:39:42 CEST] <kurosu> (for the configuration step)
[20:46:25 CEST] <__tim> what's the 45s ?
[20:49:01 CEST] <cone-679> ffmpeg 03Thomas Guillem 07master:ce2330bdf896: avcodec/videotoolboxenc: fix mutex/cond leak in error path
[20:49:02 CEST] <cone-679> ffmpeg 03Thomas Guillem 07master:9e11d27c25bf: avcodec/videotoolboxenc: split initialization
[20:49:03 CEST] <cone-679> ffmpeg 03Thomas Guillem 07master:513e6a30fb01: avcodec/videotoolboxenc: fix invalid session on iOS
[20:50:09 CEST] <cone-679> ffmpeg 03Thomas Guillem 07release/4.0:70799fae35d1: avcodec/videotoolboxenc: fix mutex/cond leak in error path
[20:50:10 CEST] <cone-679> ffmpeg 03Thomas Guillem 07release/4.0:a56eb4d56c5b: avcodec/videotoolboxenc: split initialization
[20:50:11 CEST] <cone-679> ffmpeg 03Thomas Guillem 07release/4.0:33fcbb4372e8: avcodec/videotoolboxenc: fix invalid session on iOS
[20:52:56 CEST] <January> __tim: meson
[20:54:39 CEST] <kurosu> though it miscompiles things as if it misdetected stuff related to msys2/win32
[20:55:34 CEST] <kurosu> all format specifiers to eg printf seem to be incorrectly set
[20:56:01 CEST] <kurosu> posix_memalign <- uhm, right
[20:56:42 CEST] <kurosu> so, how do I tell meson/the build system to target a windows build?
[20:57:08 CEST] <kurosu> note: not trying to learn anything for the long term, just checking what's what
[20:57:52 CEST] <nevcairiel> shouldnt it figure this out on its own, otherwise it seems a bit worthless =p
[20:58:46 CEST] <kurosu> well, I don't want to be too strict with a proof of concept
[20:58:58 CEST] <Mathieu_Du> I assume kurosu means corss-compiling ?
[20:59:12 CEST] <kurosu> certainly the people behind don't want to waste time if in the end it is wasted
[20:59:34 CEST] <kurosu> Mathieu_Du, more or less, yes, as it is Win7 x64 (MSYS2) -> Win7 x64
[20:59:49 CEST] <nevcairiel> that doesnt usually qualify as cross-compiling
[21:00:23 CEST] <kurosu> I agree, but for the sake of the build system, I don't know, hence my mentioning this
[21:00:51 CEST] <cone-679> ffmpeg 03Aman Gupta 07master:37c2cb6a68a4: Revert "avcodec/mediacodecdec: wait on first frame after input buffers are full"
[21:01:33 CEST] <cone-679> ffmpeg 03Aman Gupta 07release/4.0:789bac72eded: Revert "avcodec/mediacodecdec: wait on first frame after input buffers are full"
[21:05:33 CEST] <cone-679> ffmpeg 03Aman Gupta 07master:cd86c5dbcca5: avformat/mpegts: parse large PMTs with multiple tables
[21:06:01 CEST] <cone-679> ffmpeg 03Aman Gupta 07release/4.0:b5106c5aa2dd: avformat/mpegts: parse large PMTs with multiple tables
[21:06:26 CEST] <kurosu> #define CONFIG_DECODERS 0 <- well that looks underwhelming
[21:06:29 CEST] <jamrial> kurosu: 4m on win7 sounds about right
[21:06:34 CEST] <jamrial> wish i still had a non updated win10 machine to confirm falls creator build fucked things up
[21:06:37 CEST] <__tim> kurosu, I don't know if msys/mingw has been tested as well or if only msvc has been tested
[21:07:07 CEST] <__tim> kurosu, so might need some tweaks here and there for msys/mingw if not
[21:07:08 CEST] <kurosu> currently looking at the sorted config.h diff between configure and meson
[21:08:14 CEST] <__tim> (and that is a limitation of the port then, not of Meson)
[21:09:54 CEST] <kurosu> yeah, but I wanted an apple-to-apple comparison if at least it's the same genus but slightly different species
[21:10:03 CEST] <kurosu> here, I can't look at the compilation then
[21:10:33 CEST] <kurosu> note: I'm not even a ffmpeg user these days, so I'm really looking at this out of curiosity
[21:11:43 CEST] <kurosu> Let's try with copying over the config.*
[21:12:09 CEST] <kurosu> not working
[21:12:18 CEST] <kurosu> gcc likely incorrectly invoked
[21:12:30 CEST] <kurosu> ah well, thanks for trying
[21:45:22 CEST] <kurosu> miss these defines (at least): -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D__USE_MINGW_ANSI_STDIO=1 -D__printf__=__gnu_printf__
[21:46:24 CEST] <__tim> thanks, any chance you can chuck that it into an issue on github or so? :)
[21:46:57 CEST] <Mathieu_Du> fwiw meson will set "_FILE_OFFSET_BITS" itself
[21:47:01 CEST] <__tim> the file offset bits meson should be doing automatically already afaik, not sure why they're not picked up here
[21:47:30 CEST] <Mathieu_Du> I'm pretty sure it should be present in the ninja build file
[21:47:43 CEST] <kurosu> anyway, I need to copy over the config.* but then I need to remove manually all swscale, dxva and hwaccel
[21:47:51 CEST] <kurosu> I'm mostly running out here
[21:48:50 CEST] <kurosu> that was a messy job to try and see more of it, but at this stage, this requires going beyond what I intended to do
[21:49:05 CEST] <Mathieu_Du> What are you doing btw, cross-compiling ?
[21:49:41 CEST] <Mathieu_Du> Copying over config.h is most certainly not the right thing to do :)
[21:49:45 CEST] <kurosu> yes and no, I don't know how alien a MSYS2 compilation is - but I don't need a cross-compiler or anything, just the native one
[21:50:04 CEST] <kurosu> well, the issue is the generated config.h has too many mismatches
[21:50:21 CEST] <kurosu> so sure, some are intended, but I can't go over all of them
[21:50:24 CEST] <Mathieu_Du> Also note that you need to work in a clean checkout, some places in ffmpeg include "confg.h"
[21:50:28 CEST] <__tim> on windows it's just another native environment
[21:50:46 CEST] <Mathieu_Du> with quotes as opposed to <>
[21:51:19 CEST] <kurosu> well, to be frank, I ran out of steam for tonight
[21:51:56 CEST] <kurosu> if others managed to do make it work, then so be it
[21:51:58 CEST] <Mathieu_Du> What do you mean by "mismatches" ?
[21:52:06 CEST] <__tim> thanks for giving it a go, even if it's not something that's been tested yet :)
[21:52:26 CEST] <Mathieu_Du> It works from a Visual Studio shell ;)
[21:55:00 CEST] <kurosu> Mathieu_Du, with the proper CFLAGS hacked into build.ninja it would still misuses definitions, so likely some incorrect HAVE_ or CONFIG_
[21:56:17 CEST] <Mathieu_Du> "misuses definitions" ?
[21:56:36 CEST] <kurosu> printf format identifiers etc
[21:56:59 CEST] <Mathieu_Du> Can you paste the actual errors somewhere?
[21:57:17 CEST] <kurosu> sorted config.h differ by more than 500 lines over the one generated by configure
[21:57:32 CEST] <Mathieu_Du> That much is expected
[21:57:49 CEST] <kurosu> the problem is likely some are incorrect
[21:58:50 CEST] <Mathieu_Du> That there's a diff doesn't imply incorrectness, mostly due to the fact that the meson build definitions do not expose options
[21:58:51 CEST] <kurosu> https://pastebin.com/cFWf58wB <- example 
[21:59:03 CEST] <Mathieu_Du> And instead automatically detects everything
[21:59:10 CEST] <Mathieu_Du> There probably are some missing bits
[21:59:38 CEST] <Mathieu_Du> But copying over a config.h produced by configure will only lead to chaos :)
[21:59:46 CEST] <kurosu> another example: <SNIP>\ffmpeg\meson/../libavutil/mem.c:87: undefined reference to `posix_memalign'
[22:00:11 CEST] <kurosu> I bet, but I got to choose between chaos and chaos atm
[22:00:54 CEST] <kurosu> #define HAVE_POSIX_MEMALIGN 1 <- misdetection
[22:00:58 CEST] <nevcairiel> I would think that at least when its actually done, they would generate the same config.h, or something aint quite right
[22:01:15 CEST] <kurosu> it can't compile inline asm, so swscale has to go
[22:01:31 CEST] <kurosu> and all dxva code needs to be deactivated it seems (libs needed?)
[22:01:43 CEST] <Mathieu_Du> nevcairiel, when it's actually done yes, provided options are exposed
[22:02:14 CEST] <kurosu> The showstopper is though that kind of line: "#define CONFIG_DECODERS 0"
[22:02:41 CEST] <kurosu> why is it needed? that's likely the very thing ffmpeg is built for
[22:02:54 CEST] <Mathieu_Du> kurosu, just hacking around is bound to be pretty improductive, either report an issue and I might look at it if I find time to look at mingw, or fix the build definitions I suppose :)
[22:03:12 CEST] <Mathieu_Du> It isn't needed, and not hardcoded obviously
[22:03:51 CEST] <Mathieu_Du> And that's also not a show stopper, neither the dependency graph nor the code actually uses that confguration option ;)
[22:04:32 CEST] <kurosu> bah, as I said, I'm not going to spend more time on this - hopefully, others will
[22:05:06 CEST] <Mathieu_Du> As for the actual explanation, it can be found here: https://github.com/Alexpux/mingw-w64/blob/master/mingw-w64-headers/crt/_mingw_print_pop.h
[22:05:08 CEST] <Mathieu_Du> erm
[22:05:14 CEST] <Mathieu_Du> https://github.com/FFmpeg/FFmpeg/blob/master/configure#L3707-L3711 rather
[22:05:44 CEST] <Mathieu_Du> The dependency graph is mostly auto-generated by a script  I wrote to parse configure
[22:06:59 CEST] <Mathieu_Du> Programatically added nodes such as here "decoders, encoders, ..." are not caught by that script
[22:07:51 CEST] <Mathieu_Du> fairly easy to fix, but as I said that specific configuration bit is, to the best of my knowledge, not actually used
[22:08:52 CEST] <Mathieu_Du> Anyway, that's not a meson issue, just part of the half thousand details that still need doing in that specific port :)
[00:00:00 CEST] --- Tue Jun 19 2018

More information about the Ffmpeg-devel-irc mailing list