[FFmpeg-devel] [PATCH] avformat/framecrcenc: compute the checksum for side data

[Michael Niedermayer]

On Mon, May 27, 2024 at 04:52:22PM -0300, James Almer wrote:
> On 5/27/2024 4:50 PM, Michael Niedermayer wrote:
> > On Mon, May 27, 2024 at 04:33:21PM -0300, James Almer wrote:
> > > On 5/27/2024 4:31 PM, Michael Niedermayer wrote:
> > > > On Mon, May 27, 2024 at 09:20:55PM +0200, Michael Niedermayer wrote:
> > > > > On Mon, May 27, 2024 at 03:17:15PM -0300, James Almer wrote:
> > > > > > On 5/27/2024 3:11 PM, Michael Niedermayer wrote:
> > > > > > > On Mon, May 27, 2024 at 10:15:43AM +0200, Anton Khirnov wrote:
> > > > > > > > Quoting Michael Niedermayer (2024-04-27 02:36:23)
> > > > > > > > > This allows detecting issues in side data related code, same as what
> > > > > > > > > framecrc does for before already for packet data itself.
> > > > > > > > > 
> > > > > > > > > Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> > > > > > > > > ---
> > > > > > > > 
> > > > > > > > I am against this patch. Checksumming side data is a fundamentally wrong
> > > > > > > > thing to do.
> > > > > > > 
> > > > > > > It, or something equivalent is neccessary for regression testing.
> > > > > > > (and it was you who asked also for the tests i run to be part of
> > > > > > >     fate. But here you object to it)
> > > > > > > 
> > > > > > > You know, not checking side data is not checking it so differences would then not be
> > > > > > > detected allowing for unintended changes to be introduced (aka bugs)
> > > > > > 
> > > > > > You have seen how much code is needed to get hashing to work for all targets
> > > > > > with some types,
> > > > > 
> > > > >    framecrcenc.c |   76 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
> > > > >    1 file changed, 73 insertions(+), 3 deletions(-)
> > > > > 
> > > > > 70 more lines of code, in my patch
> > > > > 
> > > > > If we need another 70 to handle some corner cases, no idea if we do, thats
> > > > > still negligible
> > > > > 
> > > > > 
> > > > > > so it does feel like it's not the right thing to do.
> > > > > 
> > > > > I dont think i can follow that logic
> > > > > 
> > > > > 
> > > > > > ffprobe (and f_sidedata) are what should be used for actual integrity
> > > > > > checks.
> > > > > 
> > > > > ffprobe cannot test ffmpeg, ffmpeg is a seperate excutable
> > > > > 
> > > > > If you suggest that side data should not be tested in FFmpeg while packet.data
> > > > > should be tested. That position seems inconsistant to me
> > > > > 
> > > > > If you suggest that neither side data nor packet.data should be tested in FFmpeg
> > > > > iam confident that there would be a majority disagreeing.
> > > > > 
> > > > > f_sidedata is not at the output of ffmpeg so even if it could test it, it
> > > > > does not test the ffmpeg output.
> > > > > We also dont replace running md5sum and framecrc on ffmpeg output by a bitstream
> > > > > filter.
> > > > > 
> > > > > Again, there is need to test what comes out of FFmpeg, thats at the muxer level
> > > > > thats what framecrcenc does.
> > > > 
> > > > There is also an additional aspect
> > > > and that is efficiency or "time taken by all fate tests"
> > > > framecrcenc already has all the side data, it costs basically 0 time to print that
> > > > 
> > > > any ffprobe based check needs to run everything a 2nd time, so it will be slower
> > > > 
> > > > also ffprobe is only good for side data from the demuxer.
> > > > my patch tests all cases including side data from the encoder or any other
> > > > source that gets forwarded to the muxer in each testcase.
> > > 
> > > We could extend showinfo_bsf to print side data information.
> > 
> > Well, you argued a moment ago that its too much code (in framecrcenc)
> > its not going to be less code if the same or more detailed information
> > is printed in a showinfo_bsf
> > 
> > again, my suggestion is that this code should go to where side data is
> > and then showinfo_bsf, framecrcenc and ffprobe can use it
> 
> I mean, showinfo_bsf could be adapted in a way ffprobe can invoke/parse, so
> all the related ffprobe code can be moved there.

do you agree that framecrcenc should show side data in a way to allow
detecting changes, as it also does with the main packet data ?
Its perfectly fine with me if that invokes the same code as showinfo_bsf

thx

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

If a bugfix only changes things apparently unrelated to the bug with no
further explanation, that is a good sign that the bugfix is wrong.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20240530/3e923a16/attachment.sig>