[FFmpeg-devel] [PATCH v7 1/3] avutil/log: Add log flag AV_LOG_PRINT_MEMADDRESSES

[softworkz .]

> -----Original Message-----
> From: ffmpeg-devel <ffmpeg-devel-bounces at ffmpeg.org> On Behalf Of
> Andreas Rheinhardt
> Sent: Mittwoch, 9. April 2025 09:28
> To: ffmpeg-devel at ffmpeg.org
> Subject: Re: [FFmpeg-devel] [PATCH v7 1/3] avutil/log: Add log flag
> AV_LOG_PRINT_MEMADDRESSES
> 
> softworkz:
> > From: softworkz <softworkz at hotmail.com>
> >
> > which is controls prefix formatting. With this flag set, the prefix is
> > printed including the memory address, otherwise it is omitted.
> > In libavutil, the flag is set by default, retaining the previous
> > behavior. fftools remove the flag as default.
> 
> The implementation of this flag is counter to the usual one: It is
> enabled by default, but every av_log_set_flags() that does not set
> AV_LOG_PRINT_MEMADDRESSES implicitly disables printing memory addresses.

Hi Andreas,

thanks for looking at the patchset.

The intention is that memadress printing is enabled by default for library use of avutil (so that there's no change in behavior) but not for fftools (following the suggestion from Marton).

For fftools, the default is to _not_ print memory addresses _unless_ a user specifies something like:

./ffmpeg -v +memaddresses .....


> This means that it will affect users that do not opt into this. I don't
> think this is intended

Yes, it is.


> or that the new system makes sense.

Do you think it would make more sense to negate the flag semantic?


> AV_LOG_PRINT_MEMADDRESS seems better given that every av_log() will
> likely only print one memaddress.

It can be more than a single address per log line since it also prints the parent's memory address in case it specifies parent_log_context_offset.

see: https://github.com/ffstaging/FFmpeg/blob/02eda84bf2fcf0db7793872204b0f564a6557232/libavutil/log.c#L329-L337


Also, thanks a lot for the other hints!

sw