Benchmark output regulated by RMANCLIENTOPT

[Dimitar Misev]

This should either be

• documented, or
• changed to rasserver options for example

Currently I've figured from the code in source:raslib/rminit.cc, that the RMANCLIENTOPT environment variable can have several options to control the debug and benchmark levels:

• -l <logfile>: set the log file
• -d <dbgfile>: set debug file
• -dl <dbglevel>: set debug level, 0-4
• -b <bmfile>: set benchmark file
• -bl <bmlevel>: set benchmark level, 0-4. The benchmark level is 0 (off) by default, so timers are not printed irrelevant of whether --enable-benchmark was set at configure or not, unless this is set:

  export RMANCLIENTOPT="-bl 4"
  

• -notilint, -tilesize, -useTC, -timeout, -notimeout

[Peter Baumann]

hm, I find this in "./configure —help":

—enable-debug this feature enables generation of binaries that

contain extra code for debug output. The output is
usually very verbose. It is not recommended for
production use.

—without-debug-symbols compile rasdaman without debug symbols

any idea what extra trick the 2nd option does?

[Peter Baumann]

I have established a wiki page to collect all this:

​http://rasdaman.org/wiki/DebuggingBenchmarking

Once sufficiently stable we can think about migrating it into the installation guide.

[Dimitar Misev]

By default gcc gets these parameters (allows to get segfault stacktrace and run gdb to debug):

-g3 -gdwarf-2 -O0 -rdynamic

--without-debug-symbols changes that to

-O2