Closed interfect closed 8 years ago
It would be better to do 10 retries in loop and if that fails, error out with nice message.
Why does it fail in the first place?
Probably due to low entropy.
Maybe? But if it's low entropy, why does it fail the first time, and then succeed all the other times? You'd expect it to sometimes work the first time, and sometimes fail subsequent times.
On Sun, Jan 31, 2016 at 6:44 AM, Jakub Sztandera notifications@github.com wrote:
Probably due to low entropy.
— Reply to this email directly or view it on GitHub https://github.com/hyperboria/cjdns/pull/102#issuecomment-177514327.
I can only :+1: this pathc if we know what's going on
That makes sense. Unfortunately, it might take more digging into the internals of Node that I'm willing to do.
I've thrown a bunch of Node versions at the cjdns build script on this
hardware, and they generally seem to provide broken behavior. Some of them,
for example, decide that any number converted to hex is "C", while others
crash when you ask for os.cpus()
. I think this and the other similar
issues I opened might just come down to "Node is severely broken on
Angstrom on BBB", and should be reported there instead of here.
On Mon, Feb 1, 2016 at 5:57 AM, Lars Gierth notifications@github.com wrote:
I can only [image: :+1:] this pathc if we know what's going on
— Reply to this email directly or view it on GitHub https://github.com/hyperboria/cjdns/pull/102#issuecomment-177982510.
I'm just going to close this, as I really do think the problem is a badly broken node.
Sorry that we didn't merge it, but it doesn't make much sense from the project stance. You can create your fork/branch with those things fixed but I would report all those bugs to the node devs.
Fixes hyperboria/cjdns#101.