Shaka hangs on http upload if curl does not read upload cache #1347

petzeb · 2024-02-21T06:41:38Z

System info

Operating System: Ubuntu 20.04.6 LTS (Focal Fossa)
Shaka Packager Version: main, sha: 76eb2c1-release

Issue and steps to reproduce the problem

Packager Command:

packager -v=3 input=short_h264_video.mp4,stream=0,init_segment=output/init.mp4,segment_template=output/$Number%05d$.mp4 --segment_duration=10 --fragment_duration=10 --use_fake_clock_for_muxer=1 --io_block_size=65536 --time_shift_buffer_depth=50 --preserved_segments_outside_live_window=0 --mpd_output=http://localhost:43831/dash.mpd

Extra steps to reproduce the problem?
(1) Setup a local web server to receive the manifest updates, and configure the webserver to respond with an error code (403 for example) on each request.
(2) Run repeatedly until shaka hangs, it appears this issue is timing sensitive.

What is the expected result?
Shaka exits unsuccessfully

What happens instead?
Shaka hangs.

Input file is probably not important here, but for reference it was generated via:

ffmpeg -f lavfi -i testsrc=duration=10:size=1280x720:rate=30  -t 10s  sample-content/short_h264_video.mp4

Problem was originally discovered on shaka 2.6.1, attaching abbreviated verbose logs from that run (with some additional locally added logs). Issue is verified to reproduce on shaka main, but could not get the log verbosity to work well enough to extract logs there.

shaka_hangs_abbrev.log

From the logs it can be seen that, HttpFile::Flush() is entered but not exited:

stderr(372): [0220/063540:VERBOSE2:http_file.cc(265)] Flush enter

There are no logs from HttpFile::CurlReadCallback() and the main process has finished:

stderr(424): [0220/063540:VERBOSE1:status.cc(74)] 7 (HTTP_FAILURE): HTTP response code said error, response code: 403.
stderr(425): [0220/063540:VERBOSE2:http_file.cc(363)] ThreadMain about to finish
stderr(426): [0220/063540:VERBOSE2:http_file.cc(367)] ThreadMain finished

The text was updated successfully, but these errors were encountered:

joeyparrish · 2024-02-22T22:23:50Z

Thank you for the report! We will investigate.

In some cases it can happen that the http server responds with a non-successful status code without reading the response body. In this case curl may decide not to read from the cache since there is really no point in sending data to the server. In case some other thread of shaka has already called HttpFile::Flush it may end up deadlocked there waiting for the cache to either close or become empty. Thus, we close the cache when leaving the main thread as no data will be read by curl after it has finished anyways. Closes #1347

* fix: http_file: Close upload cache on task exit (shaka-project#1348) In some cases it can happen that the http server responds with a non-successful status code without reading the response body. In this case curl may decide not to read from the cache since there is really no point in sending data to the server. In case some other thread of shaka has already called HttpFile::Flush it may end up deadlocked there waiting for the cache to either close or become empty. Thus, we close the cache when leaving the main thread as no data will be read by curl after it has finished anyways. Closes shaka-project#1347 * feat: Add support for single file TS for HLS (shaka-project#934) This is based on comments at shaka-project#891. The muxer is deciding whether to write to a single file or a segment file based on the configuration. Example: ``` ../packager 'in=TOS.ts,stream=video,output=tos_video.ts,playlist_name=tos_video.m3u8' \ 'in=TOS.ts,stream=audio,output=tos_audio.ts,playlist_name=tos_audio.m3u8' \ --hls_master_playlist_output tos.m3u8 ``` Tested the content using Exoplayer. --------- Co-authored-by: Cosmin Stejerean <cstejerean@meta.com> * feat: Add PlayReady support in HLS. (shaka-project#1011) Co-authored-by: Cosmin Stejerean <cstejerean@meta.com> * feat: Add input support for EBU Teletext in MPEG-TS (shaka-project#1344) Replaces shaka-project#1181 * Add support for EBU Teletext input following Level 1.5 of the core specification ETSI EN 300 706 V1.2.1 (2003-04). * Add support for webvtt in MP4 segments output. Closes shaka-project#272 --------- Co-authored-by: Marcus Spangenberg <marcus.spangenberg@eyevinn.se> * feat: add startwithSAP/subsegmentstartswithSAP for audio tracks (shaka-project#1346) Add startwithSAP/subsegmentstartswithSAP for aac, ac3, ec3 and ac4 audio tracks according to LIVE or VOD profile. Replaces shaka-project#1055 Partial solution for shaka-project#364 --------- Co-authored-by: Xingzhao Yun <xyun@dolby.com> Co-authored-by: Joey Parrish <joeyparrish@google.com> * feat: Allow LIVE UDP WebVTT input (shaka-project#1349) An updated version of PR shaka-project#1027 That previous PR was done using 2021 code, and there were many changes in the codebase from there, so a rebase was needed and also some minor tweak here and there. But it's the same code, just reimplemented on a newer codebase. If you want to take a look at this in action, after building shaka packager with this PR's code included, try this commands in 3 different simultaneous bash sessions: 1. Video UDP input: `ffmpeg -f lavfi -re -i "testsrc=s=320x240:r=30,format=yuv420p" -c:v h264 -sc_threshold 0 -g 30 -keyint_min 30 -r 30 -a53cc 1 -b:v 150k -preset ultrafast -r 30 -f mpegts "udp://127.0.0.1:10000?pkt_size=1316"` 2. WebVTT UDP input: `for sec in $(seq 0 9999) ; do printf "%02d:%02d.000 --> %02d:%02d.000\ntest second ${sec}\n\n" "$(( ${sec} / 60 ))" "$(( ${sec} % 60 ))" "$(( (${sec} + 1) / 60 ))" "$(( (${sec} + 1) % 60 ))" ; sleep 1 ; done > /dev/udp/127.0.0.1/12345` 3. shaka packager command line: `timeout 60 path/to/build/packager/packager 'in=udp://127.0.0.1:10000?timeout=8000000,stream_selector=0,init_segment=240_init.m4s,segment_template=240_$Number%09d$.m4s,bandwidth=150000' 'in=udp://127.0.0.1:12345?timeout=8000000,stream_selector=0,input_format=webvtt,format=webvtt+mp4,init_segment=text_init.m4s,segment_template=text_$Number%09d$.m4s,language=eng,dash_roles=subtitle' --mpd_output ./manifest.mpd --segment_duration 3.2 --suggested_presentation_delay 3.2 --min_buffer_time 3.2 --minimum_update_period 3.2 --time_shift_buffer_depth 60 --preserved_segments_outside_live_window 1 --default_language=eng --dump_stream_info 2>&1` Note the added `input_format=webvtt` to the shaka packager command's second selector. That's new from this PR. If you don't use that, shaka's format autodetection will not detect the webvtt format from the input, as explained in shaka-project#685 (comment). Try the command without it if you want to. Fixes shaka-project#685 Fixes shaka-project#1017 --------- Co-authored-by: Daniel Cantarín <canta@canta.com.ar> * fix: Always log to stderr by default (shaka-project#1350) This tweaks the default config for stderrthreshold from absl/log so that we always get logs to stderr by default, as we did in v2. The --quiet and --v flags that existed in v2 can still be used to modify the log level, as well as the new --minloglevel from absl/log. Issue shaka-project#1325 * feat: Add support for ALAC codec (shaka-project#1299) Co-authored-by: Cosmin Stejerean <cstejerean@meta.com> * fix: Fix missing newline at the end of usage (shaka-project#1352) Without this, the command prompt will not appear on its own line after running `packager` with no arguments. * build: Disable warnings in third-party code (shaka-project#1353) We already disabled warnings-as-errors in third-party code, but this goes futher to completely disable third-party warnings. They are mostly noise, and distracting. If they aren't errors, we shouldn't have our build logs flooded with them. * fix: AudioSampleEntry size caluations due to bad merge (shaka-project#1354) from ALAC pull request * feat: Portable, fully-static release executables on Linux (shaka-project#1351) This adds the option FULLY_STATIC to create fully-static executables. To create portable, fully-static release executables on Linux, we need to use musl instead of glibc. Static executables from glibc are not portable. The popular musl-gcc wrapper does not support C++, so instead we use a full musl cross-compiler toolchain in the build workflow. To build FULLY_STATIC, the user must point to the appropriate cross-compiler, as we do in the workflow. On systems where musl is the native libc (such as Alpine Linux), this is not necessary. I have also read that musl's allocator is not very fast in multi-threaded applications. So when FULLY_STATIC is enabled, we will also enable mimalloc, a replacement allocator that is very fast. I tested a very basic packaging command to compare speeds of dynamic glibc, static musl, and static musl+mimalloc: dynamic glibc: runs: 2.527, 2.798, 2.703, 2.756, 2.959 avg = 2.749, std dev = 0.156s static musl: runs: 2.813, 2.920, 3.129, 3.003, 2.738 avg = 2.921s, std dev = 0.154s static musl+mimalloc: runs: 2.291, 2.034, 2.415, 2.303, 2.265 avg = 2.262s, std dev = 0.140s The mimalloc build is 82% faster than musl default allocator, 77% faster than glibc, and has more consistent runtime characteristics (lower standard deviation). * fix: Restore support for legacy FairPlay system ID (shaka-project#1357) Fixes shaka-project#1356 which was caused by the fix in shaka-project#1281 which updated this to use the correct FairPlay system ID. However since old versions recognized the previous system ID this restores support for it to avoid breaking clients. * fix: use a better estimate of frame rate for cases with very short first sample durations (shaka-project#838) Use the second sample in mp4 and webm formats. shaka-project#835 had issues with merging due to golden file conflicts. Because we cannot make dependent pull requests, this is a replica of shaka-project#835. --------- Signed-off-by: Cosmin Stejerean <cstejerean@meta.com> Co-authored-by: Cosmin Stejerean <cstejerean@meta.com> * chore(main): release 3.0.0 (shaka-project#1312) ## [3.0.0](shaka-project/shaka-packager@v2.6.1...v3.0.0) (2024-02-28) ### ⚠ BREAKING CHANGES * Update all dependencies * Drop Python 2 support in all scripts * Replace glog with absl::log, tweak log output and flags * Replace gyp build system with CMake ### Features & Bug Fixes Too many to list... See CHANGELOG.md for a full accounting! --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). * ci: Add missing arm64 (self-hosted) builds on release (shaka-project#1358) The build invocation from the release workflow needs settings to trigger self-hosted builds, including Linux arm64. * fix(CI): Add Mac-arm64 to build matrix (shaka-project#1359) Support for this came out in January, 2024. Explicit macos versions seem to be necessary for now, until GitHub offers "latest" labels targeting specific architectures. * ci: Update actions (shaka-project#1360) This silences warnings about outdated node runtimes in CI jobs. - Trivial upgrades: - actions/checkout@v3 => v4 - actions/deploy-pages@v2 => v4 - actions/upload-pages-artifact@v2 => v3 - mxschmitt/action-tmate@v3.6 => v3.17 - mikepenz/action-junit-report@v3.8.0 => v4 - Upgrades that required small changes or deeper testing: - actions/download-artifact@v3 => v4 - actions/upload-artifact@v3 => v4 - google-github-actions/release-please-action@v3 => v4 - Replaced: - svenstaro/upload-release-action@v2 => gh CLI See also: - https://github.com/actions/upload-artifact/blob/main/docs/MIGRATION.md * chore(main): release 3.0.1 (shaka-project#1361) * fix: duplicate representation id for TTML when forced ordering is on (shaka-project#1364) Fixes shaka-project#1362 * chore(main): release 3.0.2 (shaka-project#1365) :robot: I have created a release *beep* *boop* --- ## [3.0.2](shaka-project/shaka-packager@v3.0.1...v3.0.2) (2024-03-07) ### Bug Fixes * duplicate representation id for TTML when forced ordering is on ([shaka-project#1364](shaka-project#1364)) ([0fd815a](shaka-project@0fd815a)), closes [shaka-project#1362](shaka-project#1362) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). * force exact command-line ordering for HLS by breaking grouping --------- Signed-off-by: Cosmin Stejerean <cstejerean@meta.com> Co-authored-by: Peter Zebühr <33069501+petzeb@users.noreply.github.com> Co-authored-by: sr90 <sr1990003@gmail.com> Co-authored-by: Cosmin Stejerean <cstejerean@meta.com> Co-authored-by: wjywbs <wjywbs@users.noreply.github.com> Co-authored-by: Marcus Spangenberg <marcus.spangenberg@eyevinn.se> Co-authored-by: Xingzhao Yun <xyun@dolby.com> Co-authored-by: Joey Parrish <joeyparrish@google.com> Co-authored-by: Daniel Cantarín <omega_canta@yahoo.com> Co-authored-by: Daniel Cantarín <canta@canta.com.ar> Co-authored-by: Joey Parrish <joeyparrish@users.noreply.github.com> Co-authored-by: Anthony Lu <67125539+koln67@users.noreply.github.com> Co-authored-by: Shaka Bot <13107131+shaka-bot@users.noreply.github.com> Co-authored-by: Cosmin Stejerean <cstejerean@gmail.com>

petzeb mentioned this issue Feb 21, 2024

fix: http_file: Close upload cache on task exit #1348

Merged

joeyparrish added type: bug Something isn't working correctly priority: P1 Big impact or workaround impractical; resolve before feature release component: HTTP upload The issue involves HTTP upload labels Feb 22, 2024

joeyparrish mentioned this issue Feb 22, 2024

chore(main): release 3.0.0 #1312

Merged

github-actions bot added this to the v3.0 milestone Feb 22, 2024

cosmin closed this as completed in #1348 Feb 23, 2024

github-actions bot added the status: archived Archived and locked; will not be updated label Apr 23, 2024

github-actions bot locked as resolved and limited conversation to collaborators Apr 23, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Shaka hangs on http upload if curl does not read upload cache #1347

Shaka hangs on http upload if curl does not read upload cache #1347

petzeb commented Feb 21, 2024 •

edited

Loading

joeyparrish commented Feb 22, 2024

Shaka hangs on http upload if curl does not read upload cache #1347

Shaka hangs on http upload if curl does not read upload cache #1347

Comments

petzeb commented Feb 21, 2024 • edited Loading

System info

Issue and steps to reproduce the problem

joeyparrish commented Feb 22, 2024

petzeb commented Feb 21, 2024 •

edited

Loading