Re: [Bug-tar] tar: Record size = 8 blocks

[Mads Kiilerich]

Ondřej Vašík wrote:
> Ondřej Vašík wrote:
>   
> > Obviously it is change in behaviour (in tar 1.20 verbose level > 1 was 
> > required to show that warning message). Attached patch should restore
> > silence again.
> >     
>
> Sorry, I read 1.21 NEWS meanwhile and this change of behaviour is
> expected and change written in NEWS file, so obviously it is correct
> behaviour. Ignore patch ;), it seems that Mercury testsuite has to be
> fixed.
>   

For reference:
** Fixed record size autodetection.  If detected record size differs from
the expected value (either default, or set on the command line), tar
always prints a warning if verbosity level is set to 1 or greater,
i.e. if either -t or -v option is given.

But, if I understand the code and posix correctly:
* rmtread (when not _isrmt (fd) and through safe_read and safe_rw) calls 
read(2) with record_size as size_t
* it is _always_ perfectly valid for read(2) to return less than the 
requested number of bytes - "especially" when reading from a pipe, both 
also when reading from a file or tape device
* the purpose of short_read is to handle that situation and keep reading 
until a full record_size is read
* the new thing is that short_read (with -t option) will give a warning

I don't know exactly what "record size autodetection" is, but I don't 
see how it can be done before EOF has been reached. Doing it in the 
beginning of short_read does not seem right.

And it seems like there are other checks in place when EOF is reached, 
controlled by read_full_records ("Suppress fatal error on reading a 
partial record").

/Mads