Hylafax on FreeBSD 10 and 100% CPU

[Konstantin Belousov]

On Thu, Jul 03, 2014 at 10:44:34AM +0100, Daren wrote:
> On 02/07/2014 23:24, Konstantin Belousov wrote:
> > On Wed, Jul 02, 2014 at 10:35:16AM +0100, Daren wrote:
> >> Hi
> >>
> >> I've tried installing Hylafax from ports on fresh FreeBSD 10.  We
> >> previously had it running on an 8.2 system without issue.
> >>
> >> After starting faxgetty creeps up to and stays on 100% cpu.  I did find
> >> a previous PR for this
> >> (https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=166071) which has the
> >> status of resolved fixed although people have requested it be re-opened
> >> as the problem is still occurring for them as well.
> >>
> >> I can add another comment to this if it helps get it noticed again, but
> >> my knowledge with C etc is non-existent to be able to help with it,
> >> although searching has come up with a similar/same issue on dragonflybsd
> >> (http://bugs.dragonflybsd.org/issues/2028)
> >>
> >> Does anyone know if this is being looked at?
> >>
> > 
> > I once looked at the source code of the program, and the ktrace reports
> > in the mentioned PR 166071 are consistent with what I remember I saw in
> > code, as well as what was discussed at that time.
> > 
> > If you look at the kdump, you note the series of read(2) syscalls on the
> > same FIFO, which return 0, the indicator of the EOF.  AFAIR, the Hylafax
> > code does  not check for the EOF condition and just spins trying to
> > read more data.  Select(2) returns the fd for EOF'ed FIFO ready, because
> > the read(2) indeed does not block at EOF.
> > 
> > My decision at that time was that the issue is the program bug.
> > I do not use the program, and cannot set it up locally to even try
> > coding the fix.
> > 
> 
> Thanks for the answer.  Unfortunately I know nothing about these sorts
> of things so it doesn't mean much to me!
> 
> I'll try posting to the hylafax list with this, but I'm sure I came
> across this previously mentioned to them on a search and they made out
> it was an OS bug - although now I cannot find it again.

I am happy to listen to the authors of this application, in particular,
I am interested in exact description of what behaviour do they expect.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-ports/attachments/20140703/9755b785/attachment.sig>