[FFmpeg-devel] [PATCH v4 06/10] libavformat/asfdec: fix macro definition and use

Andreas Rheinhardt andreas.rheinhardt at outlook.com
Mon May 16 11:48:57 EEST 2022


Soft Works:
> 
> 
>> -----Original Message-----
>> From: ffmpeg-devel <ffmpeg-devel-bounces at ffmpeg.org> On Behalf Of
>> Andreas Rheinhardt
>> Sent: Sunday, May 15, 2022 8:12 PM
>> To: ffmpeg-devel at ffmpeg.org
>> Subject: Re: [FFmpeg-devel] [PATCH v4 06/10] libavformat/asfdec: fix
>> macro definition and use
>>
>> softworkz:
>>> From: softworkz <softworkz at hotmail.com>
>>>
>>> Signed-off-by: softworkz <softworkz at hotmail.com>
>>> ---
>>>  libavformat/asfdec_f.c | 24 ++++++++++++------------
>>>  1 file changed, 12 insertions(+), 12 deletions(-)
>>>
>>> diff --git a/libavformat/asfdec_f.c b/libavformat/asfdec_f.c
>>> index 81a29f99d5..91c3874ac7 100644
>>> --- a/libavformat/asfdec_f.c
>>> +++ b/libavformat/asfdec_f.c
>>> @@ -906,21 +906,21 @@ static int asf_read_header(AVFormatContext *s)
>>>  }
>>>
>>>  #define DO_2BITS(bits, var, defval)             \
>>> -    switch (bits & 3) {                         \
>>> +    switch ((bits) & 3) {                       \
>>>      case 3:                                     \
>>> -        var = avio_rl32(pb);                    \
>>> +        (var) = avio_rl32(pb);                  \
>>>          rsize += 4;                             \
>>>          break;                                  \
>>>      case 2:                                     \
>>> -        var = avio_rl16(pb);                    \
>>> +        (var) = avio_rl16(pb);                  \
>>>          rsize += 2;                             \
>>>          break;                                  \
>>>      case 1:                                     \
>>> -        var = avio_r8(pb);                      \
>>> +        (var) = avio_r8(pb);                    \
>>>          rsize++;                                \
>>>          break;                                  \
>>>      default:                                    \
>>> -        var = defval;                           \
>>> +        (var) = (defval);                       \
>>>          break;                                  \
>>>      }
>>>
>>> @@ -1003,9 +1003,9 @@ static int asf_get_packet(AVFormatContext *s,
>> AVIOContext *pb)
>>>      asf->packet_flags    = c;
>>>      asf->packet_property = d;
>>>
>>> -    DO_2BITS(asf->packet_flags >> 5, packet_length, s-
>>> packet_size);
>>> -    DO_2BITS(asf->packet_flags >> 1, padsize, 0); // sequence
>> ignored
>>> -    DO_2BITS(asf->packet_flags >> 3, padsize, 0); // padding length
>>> +    DO_2BITS(asf->packet_flags >> 5, packet_length, s->packet_size)
>>> +    DO_2BITS(asf->packet_flags >> 1, padsize, 0) // sequence
>> ignored
>>> +    DO_2BITS(asf->packet_flags >> 3, padsize, 0) // padding length
>>>
>>>      // the following checks prevent overflows and infinite loops
>>>      if (!packet_length || packet_length >= (1U << 29)) {
>>> @@ -1066,9 +1066,9 @@ static int
>> asf_read_frame_header(AVFormatContext *s, AVIOContext *pb)
>>>      asf->stream_index     = asf->asfid2avid[num & 0x7f];
>>>      asfst                 = &asf->streams[num & 0x7f];
>>>      // sequence should be ignored!
>>> -    DO_2BITS(asf->packet_property >> 4, asf->packet_seq, 0);
>>> -    DO_2BITS(asf->packet_property >> 2, asf->packet_frag_offset,
>> 0);
>>> -    DO_2BITS(asf->packet_property, asf->packet_replic_size, 0);
>>> +    DO_2BITS(asf->packet_property >> 4, asf->packet_seq, 0)
>>> +    DO_2BITS(asf->packet_property >> 2, asf->packet_frag_offset, 0)
>>> +    DO_2BITS(asf->packet_property, asf->packet_replic_size, 0)
>>>      av_log(asf, AV_LOG_TRACE, "key:%d stream:%d seq:%d offset:%d
>> replic_size:%d num:%X packet_property %X\n",
>>>              asf->packet_key_frame, asf->stream_index, asf-
>>> packet_seq,
>>>              asf->packet_frag_offset, asf->packet_replic_size, num,
>> asf->packet_property);
>>> @@ -1144,7 +1144,7 @@ static int
>> asf_read_frame_header(AVFormatContext *s, AVIOContext *pb)
>>>          return AVERROR_INVALIDDATA;
>>>      }
>>>      if (asf->packet_flags & 0x01) {
>>> -        DO_2BITS(asf->packet_segsizetype >> 6, asf-
>>> packet_frag_size, 0); // 0 is illegal
>>> +        DO_2BITS(asf->packet_segsizetype >> 6, asf-
>>> packet_frag_size, 0) // 0 is illegal
>>>          if (rsize > asf->packet_size_left) {
>>>              av_log(s, AV_LOG_ERROR, "packet_replic_size is
>> invalid\n");
>>>              return AVERROR_INVALIDDATA;
>>
>> While protecting macro arguments is good, it is not really a "fix"
>> unless current usage is buggy.
> 
> Ok, I will rephrase the commit message.
> 
>> Which it isn't here, because >> has higher precedence than &.
> 
> Could you explain which change you are referring to?
> 

Putting "bits" in parentheses. It doesn't change anything, because >>
has higher precedence than &.

> All this patch does is to put macro variables in brackets
> and remove semicolons..
> 
>> Furthermore I am not really sure whether removing the ';' is even
>> something worthwhile; they are surely unnecessary (being null
>> statements), but does this matter?
> 
> It causes a warning
> 
> https://releases.llvm.org/13.0.0/tools/clang/docs/DiagnosticsReference.html#wextra-semi-stmt
> 

I don't receive this warning despite using Clang 13.0. Do you have -Wall
or -Wextra or something like that enabled?
IMO a better fix for this would be to wrap the macro in a do {} while
(0) to keep the macro calls function-like.
Anyway, you should have mentioned in the commit message that your aim is
to fix this uncommon warning.

> I don't know how others are working, but I use to work in a way where
> such warnings are shown in the editor and in lists in the IDE
> even without compilation. Now - when you have a code file that
> generates like 20, 50 or more warnings, it's much harder to spot
> those warnings that might be really relevant and hinting at a mistake,
> and you might be just too lazy to go through them each time.
> 
> The clang diagnostics have been helpful in spotting some actual 
> issues in this very file. That's why I consider it worthwhile
> to also eliminate such "non-issues".
> 

I also work like that; e.g. my recent ac3.h header patchset was inspired
by clangd not liking cycles in header inclusions ("In included file:
main file cannot be included recursively when building a preamble").

- Andreas


More information about the ffmpeg-devel mailing list