Re: Victory! GNU Pth built / tested on Hurd!
Ah, but the Hurd servers are already built with cthreads
which is wired right into gnumach. Any OS-critical program
should rightly be built using the facilities provided by Mach,
I think, unless that is to be somehow abstracted? My understanding
was that rather than a big monolithic kernel, the idea was a bunch
of small servers easily tailored to the current microkernel. Since
we have gnumach ( and thusly cthreads ) we should use those
for any mission critical apps that require the time and effort
to ensure the maximum possible performance.
Someday when Hurd leaps to a new microkernel, all of the
Hurd servers will have to be modified to take this into account,
and the c-library will need to be updated as well, but that is
it, because no Hurd app ( other than a server ) should really be
mucking about with Mach... if it is then it is more or less cheating
and the price will eventually have to be paid.
Thomas Bushnell, BSG wrote:
Good deal. From what I was reading about pth and
its support for pthreads, it is important to note that
pth is non-preemptive. Apparently there is nothing
in the pthread API that requires this, but some apps
which expect this behavior may not function correctly.
In several key areas, the Hurd servers do require preemption.