Home | History | Annotate | Download | only in pthread_cancel
      1 This file contains various information.
      2 
      3 Please refer to http://nptl.bullopensource.org/phpBB/ for general information
      4 and questions. Feel free to post your questions there.
      5 
      6 
      7 We are doing our best to ensure the test case will be very portable.
      8 Please report any problems on the forum (see on top of this file)
      9 
     10 
     11  * Flags
     12 You may want to add -DVERBOSE=2 to have verbose tests,
     13 or -DVERBOSE=0 to have silent tests (for batchs for example).
     14 
     15 You may want to add -DSCALABILITY_FACTOR=X, where X is an integer,
     16 to change the stress programs load (default is 1).
     17 
     18 
     19  * Commands
     20 Compilation under linux:
     21 gcc -o <bin> -lpthread <source>
     22    where <bin> is the executable you want to build and <source> is the source file.
     23 
     24 Compilation under AIX5L
     25 cc -o <bin> -lpthread <source>
     26   same comment as above
     27 
     28 Compilation under Solaris 9 (gcc: http://www.sunfreeware.com)
     29 gcc -std=gnu99 -lpthread -lrt  -o <bin> <source>
     30   same comment as above.
     31   Note: the -lrt is necessary for test cases using semaphore feature.
     32 
     33  * Execution
     34 
     35 Please consider the following:
     36 -> I am sizing the tests to fit on a bi-xeon IA32 machine, with 2GB RAM.
     37 You can change the SCALABILITY_FACTOR value to 2 or 3 to be more
     38 resources consuming on a bigger system. This will create more threads,
     39 etc... according to each case.
     40 
     41 -> Stress tests will run until they are killed with SIGUSR1 (or they
     42 fail). Currently, I do
     43 $> ./stress &
     44 and then after a while
     45 $> kill -USR1 <pid>
     46 or
     47 $> for PR in `ps o pid --no-headers -Cstress`; \
     48    do echo Sending SIGUSR1 to pid $PR...; \
     49        kill -USR1 $PR; \
     50    done;
     51 Some cases will keep on executing ~ 1 minute after they receive the
     52 signal; it is normal (time for stopping all threads).
     53 
     54