[ts-gen] segmentation faults

[R P Herrold]

On Thu, 21 Aug 2008, Paul C wrote:

> Although that's with the TWS and mysql server running on a 
> different machine (the amd64) - in case that matters.

It does not.  At least it _better_ not.  Network socket 
readiness is crucial for the shim, scalability, failover, and 
so forth.

That kind of setup is part of my standard setup:
 	- the TWS lives on a OS/X ppc
 	- the database lives on a 32 or a 64 bit CentOS box
 		(Bill connects from 32 to 64 bit; I connect
 		from all others [more in a moment] to 32 bit)
 	- 'all others' is 32 and 64 bit CentOS 4 and 5; 32 bit
 		Debian, PPC OS/X

I've been working on getting more 'others' through xen 32 and 
64 bit Debian installs working the last few weeks, so we can 
replicate all the various environments we have reports from.

NetBSD under xen will then follow.

Last weekend, I refreshed my tests against Windows hosts to 
see how the Cygwin 1.7 beta (which will have the interfaces we 
require) is coming, but it is still pretty rough.  Getting xen 
installations under Win 2000, Win XP Pro, and Win Vista 
(unsure which sub-variant) ot Cygwin is wanted by me, but 
until Cygwin gets further along, we cannot test, and so are 
not so interested in that space.

Automated building, and testing scripts fire off to test on 
'all others' whenever Bill does an external release, and I 
pick at the edges to keep everything building and passing 
tests (both the ones in the external releases, and interally 
written cases) on internals. ;)

It is not a guarantee, but we try to keep all the 'four 
letter' scripts in ./exs/ working at all times;  risk.rb is 
more volitle, depending on what we are developing when an 
external gets 'cut' and pushed.

-- Russ herrold