Bug#831449: lintian on Sid amd64 reports volatile false spelling errors in binaries

["Thomas Schmitt" <scdbackup@gmx.net>]

Hi,

strings wrote:
> >  ings of E1
> >  length 41
> >  or lengH

Jakub Wilk wrote:
> This is bizarre. Do you know why GCC does it? :-/

Not at all.
I wonder how the machine code puts these fragments together to a
human readable text. But on the other hand i did not do assembler
since the days of the 6502 CPU.

The disease is present at the start of the .so file, after some
unmangled program symbols have been shown by "strings". At some
point of sequential reading the texts suddenly become normal.
The disease does not come back until the end.

Same with the GNU xorriso binary, a static compilation of 4 library
sources and a tiny main program.
But i do not see this pattern with big binaries in /usr/bin (vim.basic,
gcc-5, gpg).

Is this some special amd64 trick which is applied only at low
addresses ?
The first full length strings appear at byte addresses
  0x0aa960  of libisoburn.so
  0x104a20  of GNU xorriso
  

As said, it also happens on my Debian 8 but with less non-text bytes
put inbetween the text snippets.


Have a nice day :)

Thomas