mobical - Man Page

Model Based Tester for CaDiCaL


usage: mobical [ <option> ... ] [ <mode> ]

where '<option>' can be one of the following:


| -h    print this command line option summary and exit


print CaDiCaL's three character version and exit


print build configuration


increase verbosity


force colors for both '<stdout>' and '<stderr>'


disable colors if '<stderr>' is connected to terminal


assume '<stderr>' is not connected to terminal


do not print seeds in random mode


specify the number of solving phases explicitly

--time <seconds>

set time limit per trace (none=0, default=10)

--space <MB>

set space limit (none=0, default=1024)


consider out-of-time or memory as error


generate small formulas only


generate medium sized formulas only


generate big formulas only

Then '<mode>' is one of these


generate and execute trace for given 64-bit seed


<output>  generate trace, shrink and write it to file

<input> <output>

read trace, shrink and write it to output file


read and replay the specified input trace

The output trace is not shrunken if it is not failing.  However, before it is written it is executed, unless '--do-not-execute' is specified:


just write to '<output>' without execution

In order to check memory issues or collect coverage you can force execution within the main process, which however also means that the model based tester aborts as soon a test fails


execute all tests in main process directly

In order to replay a trace which violates an API contract use


To read from '<stdin>' use '-' as '<input>' and also '-' instead of '<output>' to write to '<stdout>'.

Implicitly add 'dump' and 'stats' calls to traces:


| -d      force dumping the CNF before every 'solve'

--stats | -s

force printing statistics after every 'solve'

Implicitly add 'configure plain' after setting options:

--plain | -p

Otherwise if no '<mode>' is specified the default is to generate random traces internally until the execution of a trace fails, which means it produces a non-zero exit code.  Then the trace is rerun and shrunken through delta-debugging to produce a smaller trace.  The shrunken failing trace is written as 'red-<seed>.trace' to the current working directory.

The following options disable certain parts of the shrinking algorithm:











The standard mode of using the model based tester is to start it in random testing mode without '<input>', '<seed>' nor '<output>' option. If a failing trace is found it will be shrunken and the resulting trace written to the current working directory.  Then the model based tester can be interrupted and then called again with the produced failing trace as single argument.

This second invocation will execute the trace within the same process and thus can directly be investigated with a symbolic debugger such as 'gdb' or maybe first checked for memory issues with 'valgrind' or recompilation with memory checking '-fsanitize=address'.


March 2024 mobical 1.9.5