search

bblum / landslide-simics

the landslide codebase (old simics version)
11 stars 9 forks source link

Fix "if (ls->test.start_kern_heap_size > ls->kern_mem.heap_size)" #181

Open bblum opened 9 years ago

bblum commented 9 years ago

Should be < not >.

This basically means this FAB check has been deadcode for the past N years. So it's not actually clear if it should be fixed (ie enabled); it might trigger way too many times on kernels that are ok.

Needs to be tested on pebbles to see if the fixed version causes false positives, or if the check can safely be enabled.

bblum commented 9 years ago

In pintos, enabling this check causes false leak reporting EVERYTIM because a dying thread's stack is freed in thread_schedule_tail(), which is after the test-case-ended trigger condition (tell_landslide_thread_switch() to idle).

For example http://www.contrib.andrew.cmu.edu/~bblum/landslide-trace-1437445937.83.html