Pierangelo Masarati skrev:
Howard Chu wrote:
The long fixed-length waits in the syncrepl tests always annoy me. Is anyone actually testing on hardware that takes longer than 1 second to sync? (The last time I tested on such a slow machine was the z/OS partition I ported to back in 2002/2003.) Would anyone object to dropping the all syncrepl intervals to say, 2 seconds, and setting the delays to around 5 seconds?
In general, no. The only case I needed long(er) sleep times was occasionally when a process needs to start under valgrind. Execution of few operations is not usually slowed down. Only heavy concurrent load shows some difference.
Why not define the sleep times as a variable in defines.sh, allowing it to be overridden by an environment variable? That should allow those that requires longer sleep times easy access to it without forcing everybody to wait that long..
Rein