[Tor-BSD] BSDCan BoF
george at ceetonetechnology.com
Tue Mar 31 20:47:47 EDT 2015
>> On 1 Apr 2015, at 06:53, George Rosamond <george at ceetonetechnology.com> wrote:
>> I pinged Dan about having a BoF at BSDCan in June, and we should confirm
>> it now.
>> We'd have about 45" during lunch, which I think is adequate.
>> If we need 'material', I can update/rework the short intro I did at
>> vBSDCon, but there's some interesting stuff happening in Tor/BSD
>> land.... buildbot, attila's work on TBB for OpenBSD...
> These changes are somewhat peripheral, but they do affect BSD more than some other platforms:
> Performance on BSD etc.
> Recent performance improvements reduce redundant time system calls, which particularly affects BSD, OS X, and Linux without virtual system calls enabled (e.g. IPredator's reported issues with 3.x kernel series under high load).
> Along with the general performance improvements in tor 2.6, this should allow tor to push more data on BSD, and reduce the observed performance gap with Linux.
> clang sanitizer undefined behaviour clean
> For those who build tor with clang (common on BSD & OS X), tor is undefined behaviour clean on 64-bit platforms according to the clang sanitizer. (There's a patch for unsafe signed left shifts in the donna code for 32-bit platforms, but it's stalled on upstream.) The clang checks complement the existing coverity checks, and help ensure tor's behaviour is well-defined whichever compiler is used.
good stuff, and all, but the topic is the BoF at BSDCan.
Who is going? Is there interest on this list doing a related BoF?
More information about the Tor-BSD