On Thu, Nov 30, 2006 at 01:44:41PM -0800, Sander Temme wrote: > So, no regression. > > However, if I build and test from a source tree in /tmp, like you, > testlfs fails for me even more verbosely than for you: > > [EMAIL PROTECTED]:/tmp/apr-1.2.8/test$./testall -v testlfs > testlfs : |Line 79: truncate file to 8gb (28): No space > left on device
That all makes sense then, definitely no need to hold up any release for these. I don't see much motivation to tweak testlfs to avoid that failure cause. I've tweaked testsockets to give a useful error message for that failure cause if somneone wants to re-run that, doubtless it is failing simply because this is a configuration which allows creating a PF_INET6 socket but refuses to bind() it with no interfaces configured. The poll issue is presumably something funky in the port-based poll backend and is Solaris-specific. Regards, joe
