|
From: Tom H. <to...@co...> - 2011-10-10 08:37:09
|
On 10/10/11 07:25, Bart Van Assche wrote: > Any chance you can have a look at the runtime failure of err_disable4 on > bristol and change the stack size into something that does not cause a > failure ? Well it's not clear there is any such value... Basically it is failing at about thread 449 with EAGAIN and the actual stack size does seem to affect that much. Even massive increases in the requested stack only reduce that failure point by one or two. Given that the requested stack size is already close to the minimum there isn't any real scope for decreasing it further. My suspicion is that it is some sort of internal glibc limit anyway as the hardware (and kernel) is exactly the same (x86_64 with 24Gb of memory) as for the other, later, Fedora versions that are working. Tom -- Tom Hughes (to...@co...) http://compton.nu/ |