Created attachment 160321 [details]
unified diff

1. The use of gdb at the end of run-assert-msg-test.sh fails on Windows,
   and doesn't seem very portable anyway.  It also fails to test
   functionality, instead testing implementation.  I commented it out.

2. The diff in run-markup-tests.sh needs --strip-trailing-cr to pass
   on Windows.

3. "gobject/ performance" can fail with a division by zero if the timers
   can't resolve a fine enough time and report 0 time elapsed.  Check for
   this condition and skip the test if it happens.

On the subject of (3), I'm running XP SP3 in a VMware virtual machine, so
that might impact the behavior of the timers.  In any event, running
"testglib -m slow --verbose" reports (in part):

===============================================================

GTest: run: /testglib/Timers (slow)
checking timers...
  spinning for 3 seconds...
ok
checking g_timer_continue...
        run for 1 second...
        stop for 1 second...
        run for 2 seconds...
        stop for 1.5 seconds...
        run for 0.2 seconds...
        stop for 4 seconds...
        run for 5.8 seconds...
        => timer = 9.015625 = 9.015625 (should be: 9.000000) (0.015625 off)
g_timer_continue ... ok

        => timer2 = 15.515625 = 15.515625 (should be: 15.500000) (0.015625 off)

timer2 ... ok

GTest: result: OK

===============================================================

I'll just note that 0.015625 = 2^-6.  Don't know if that's significant.