Emergency UNIX and Linux Support

We have a situation where a binary from our build system is failing in one specific flow but the same from dev works fine. We are sure that build system picks up the right libs and code, however we see a difference of 6000 bytes in both. (build - 6000 bytes = size of dev binary)

Hence we would like to know:
a. How we can compare two Unix Binaries when they differ only in terms of build boxes (ex dev and build)
b. Given a binary (not compiled with debug options) what are the information we can gather from it.

Thanks in advance.

Please can you post information on the O/S, version and architecture ?

O/S & Version - Red Hat Enterprise Linux Server release 5.4
Architecture - Linux xxxxx 2.6.18-164.el5 #1 SMP Tue Aug 18 15:51:48 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux

I would compare the build flags (debug builds often can "null" statically allocated memory whilst optimized builds do not)
I would do an "ldd" on the binary and find which libraries it is using.
Then for each of the libraries you can either use "rpm" to find the version installed, or use "sum" or "md5sum" to check they are the same on both hosts.
After all that I would "strace" the program running on both and compare the two outputs - this usually gives you an indication of where to start looking for failures.
I hope this helps...

Sometime, define of file block are different, so same file will show different size in different servers.

Hi.

The utility nm can extract linkage symbols from object files and executables (in most situations, e.g. where the executable has not been stripped).

You could use the lists from each version of the executable (or constituent object files) to compare for differences, although this would be at a large granularity.

See the recent post Fortran dependency checking for a use of nm ... cheers, drl

A diff of 6000 bytes, itself tells that either the code or the lib used is not exactly same at both the build locations. You mentioned that you are confirmed that these things are same, how did you arrived at the conclusion?

---------- Post updated at 08:33 PM ---------- Previous update was at 08:27 PM ----------

Great suggestion!!! I'd have done similarly, eventhough a bit cumbursome, but bound to get results. Also the version control of the source should tell what's different in the recent time, should help narrow down the suspected libraries.

---------- Post updated at 08:48 PM ---------- Previous update was at 08:33 PM ----------

Lastly, I'd also have made sure, as a first step, that the build system itself does not have different versions. Apart need to make sure that the code does not have too much of flag based compilations like:

Code:

   #ifdef <flag>
      ...
   #else
      ...
   #endif <flag>

// OR
   #ifndef  xxxx
      ...
   #endif

If yes, then I'd also suggest to have a look at the environment variables in addition to the makefile flag definition.

Good Luck.