Re: differences in binaries

Re: differences in binaries

[tmoran]

>When doing a byte-by-byte comparison of seemingly identical binaries
>(ie compile, move the resulting binary to binary.old, then compile again and compare)
>there seems to be all sorts on differences between the two files, even though
>they were only just re-linked.
>We are using SunAda 1.1 but I'm sure this occurs with other versions of Ada.
  I just tried this re-linking with Janus Ada for Windows.  It has three
steps:  compile, bind the Ada parts into a .obj file, use Microsoft's
linker to add windows stuff producing a .exe file.  A re-bind produced an
identical binary .obj file.  A relink produced an .exe that differed in
two bytes, widely separated in the .exe file.
  I don't know what might be happening in SunAda, but it clearly doesn't
'occurs with other versions of Ada'.

differences in binaries

[Craig Moore]

I have had something interesting pointed out to me about Ada which the 
collective wisdom of this newsgroup may be able to explain.

When doing a byte-by-byte comparison of seemingly identical binaries
(ie compile, move the resulting binary to binary.old, then compile again and compare)
there seems to be all sorts on differences between the two files, even though 
they were only just re-linked.

I guess it is just timestamp information but can anyone point me to some
doco's that possibly explain what the differences are ??

We are using SunAda 1.1 but I'm sure this occurs with other versions of Ada.


Thanks,

Craig.

Re: differences in binaries

[Robert S. White]

Craig Moore wrote:

> When doing a byte-by-byte comparison of seemingly identical binaries
> (ie compile, move the resulting binary to binary.old, then compile
> again and compare) there seems to be all sorts on differences 
> between the two files, even though they were only just re-linked.
> 
> I guess it is just timestamp information but can anyone point me to
> some doco's that possibly explain what the differences are ??
> 
> We are using SunAda 1.1 but I'm sure this occurs with other 
> versions of Ada.

   When I use a version of DDC-I Ada 83 and run a tool on the linker 
 output to prepare the result for ROM programming, I can get exact 
 compares of load modules.  This works if the command file for the 
 linker lists the object modules in the same order and the 
 elaboration code calls all of the component elaboratiions in the 
 same order.

   Changing a Ada spec can cause a re-build of just the neccessary 
 modules to have a different command file to the linker and different 
 elaboration code than if one deleted all object and other 
 intermediate files and did an entire rebuild from scratch.  But I 
 have seen mulitiple rebuilds from scratch generate identical code 
 for ROM programming for the particular compiler I use for some 
 target systems.  Your mileage may vary.  

   Previous threads on this subject about a year ago talked about 
 different host computer resources causing modules to be built in 
 different orders if I recall correctly.  DEC Ada?
_______________________________________________________________________
Robert S. White                        | Opinions expressed are my own.
whiter@crpl.cedar-rapids.lib.ia.us     | An Embedded Systems programmer