On 02/03/2012, at 8:18 PM, Leif Hedstrom wrote: > On 3/2/12 8:29 PM, Whitaker, James X. -ND wrote: >> >> I imagine that it isprobably humanly impossible to see THIS bug with code >> inspection.Or even by stepping through it in a debugger. >> > > > Hi James, welcome to hell... I mean, ATS development. > > I need to read this a few times I think. But before we go crazy, replacing > freelist etc., lets try to figure out how to fix it :). Maybe the first step > is to eliminate some of those zillion ifdef's. TS_MICRO is a "dead" product, > as far as I'm concerned at least. It was intended for small embedded devices > etc., and probably is completely irrelevant these days (but, jplevyak would > know best). Some of the other defines might also be unnecessary.
James, here's how I'd recommend we proceed ... First, let's open bugs to remove the TS_MICRO cruft and to track the free list issue you have described, https://issues.apache.org/jira/browse/TS. Next, I think it's really important to have a test harness that exposes the issue (at well as to measure the performance of any changes). This kind of thing can often be platform and toolchain-specific, so it's easier to get people to test across platforms with a smaller harness. With a small test case it is easier to validate that the expected machine code is being generated. cheers, James > > I'm just back from a work trip, and also have some other "commitments" to > deal with. But, hopefully we can get some momentum going on this problem too. > Please make sure to file bugs etc. > > Also, this probably belongs on dev@ and not users@. I'm Cc:both for this > reply, but I think you should also forward the original email to dev@ (and > make sure you subscribe to it if you haven't already). The future discussions > on this problem definitely belongs on dev@. > > Cheers, > > -- Leif >
