> > We understand the concern about testing times. To choose a middle > > ground, Ted's suggestion of using _scratch_mkfs_sized works best for > > CrashMonkey specific tests. These tests involve very few files and it > > suffices to have a 100MB file system. I tested the patch on ext4, xfs, > > btrfs and f2fs on a partition of this size. The overhead due to > > _check_scratch_fs after each sub test is in the range of 3-5 seconds > > 3-5 second per invocation of _check_scratch_fs? No. This is the overall overhead for calling _check_scratch_fs after each of the 37 sub tests in my first patch. Without this check, it took 10 seconds to run. With this addition, it takes about 12-15 seconds to run all the 37 sub tests. > > for all these file systems. If this is tolerable, we can force a > > Not if your numbers that means 300 x 3-5 seconds. That's 15-25 > minutes of extra runtime. Going by the above calculation, the extra run time for the invocation of check after each sub-test should be around 30-40 seconds in total (for the set of 300 tests).
