Date: Fri, 11 Jul 2014 14:19:57 +0100 From: Daren <daren.russell@ksm.co.uk> To: Konstantin Belousov <kostikbel@gmail.com>, ports@freebsd.org Cc: Lee Howard <faxguy@howardsilvan.com>, hylafax-users@hylafax.org Subject: Re: [hylafax-users] Hylafax on FreeBSD 100% CPU Message-ID: <53BFE47D.4030203@ksm.co.uk> In-Reply-To: <20140711092202.GZ93733@kib.kiev.ua> References: <53B5329F.9000906@ksm.co.uk> <53B5F55A.20909@howardsilvan.com> <20140705103816.GP93733@kib.kiev.ua> <53B8601E.8040600@howardsilvan.com> <53BD1337.7040808@ksm.co.uk> <53BD9B43.5010604@howardsilvan.com> <53BE4EA9.5000308@ksm.co.uk> <20140710083923.GK93733@kib.kiev.ua> <53BF38D7.4030402@howardsilvan.com> <20140711092202.GZ93733@kib.kiev.ua>
next in thread | previous in thread | raw e-mail | index | archive | help
Ah, you beat me to posting the findings! I thought I'd better make the point that the changes Lee made were against Hylafax+ (5.5.5 from Sourceforge) whereas the port uses Hylafax 6 from hylafax.org. I do not know 100% if the changes Lee made to H+ will filter back to hylafax.org. I was going to have an attempt at making a port locally using the H+ sources, and if successful can submit it, but unfortunately I don't know if I'll have time to do it full-time (assuming I'm successful - I've never tried anything like this before!) In the meantime, anybody having the 100% cpu issue can enable an option in the config.site file of the existing port, which is CONFIG_OPENFIFO="O_RDWR". A simple patch file added to the files directory is probably easiest (even I managed to do that!!) I've been running with this single change since this morning with no issues. This more or less falls in line with the final comment of the original PR, but it seems you do not need the FIFOBUG switch enabled, nor the Linux procfs mounted as was also in the original PR. Regards Daren On 11/07/2014 10:22, Konstantin Belousov wrote: > There was quite good and encouraging progress made on the state > of the Hylafax for the FreeBSD, thanks to the efforts of Daren and Lee > Howard. I do not know when the next release of Hylafax come out with the > committed changes below, but whenever it happens, the port update to new > release should improve things. > > The port accumulated quite a bit of problems, and most pressing issues > on FreeBSD with the 100% CPU time spent on FIFO reads, as well as > utmp->utmpx conversion, are fixed in upstream now. > > There is no maintainer for the Hylafax port, I am forwarding this > message to ports@ in hope some motivated ports person will pick the > stuff. My belief is that the programming issues are fixed, the > remaining work is packaging-related. > > On Thu, Jul 10, 2014 at 06:07:35PM -0700, Lee Howard wrote: > > On 07/10/2014 01:39 AM, Konstantin Belousov wrote: > >> On Thu, Jul 10, 2014 at 09:28:25AM +0100, Daren wrote: > >>> I had a quick go with your patch, but it didn't apply. Having a quick > >>> look I believe it's simply an issue with the path to the hfaxd folder, > >>> so I'll have another look as soon as I get a little time. I did also > >>> have a quick look at compiling version 5.5.5 (at first without the > >>> patch) but ran into issues straight away. When compiling > GettyBSD.c++, > >>> it's having a fatal error 'utmp.h' file not found. A quick google > >>> suggests this was changed to "utmpx.h" in recent versions of > FreeBSD and > >>> looking at the ports version of hylafax6, there's a few patches that > >>> need applying for that to compile properly (and although I can see > what > >>> they do, I'm unsure why they do it!) I'll try and have another go > when > >>> I get a spare hour or two at work. > >> This in fact means that FreeBSD port-specific patches should be > upstreamed. > >> But the ports does not have a maintainer. Somebody needs to > communicate > >> the patches to the Hylafax developers. > > > > I've taken what little of those patches seemed appropriate for upstream > > application and I've committed them: > > > > http://sourceforge.net/p/hylafax/HylaFAX+/2336/ > > http://sourceforge.net/p/hylafax/HylaFAX+/2337/ > > > > However, the real difference had to do with how the package in FreeBSD > > ports was being built: "--with-GETTY=SysV" instead of the "BSD" default > > that would get picked-up normally. Since the SysVGetty supported utmpx > > this worked. > > > > Nonetheless, I've updated the BSDGetty code in HylaFAX+ to support > > utmpx. So it can now be built with SysV or BSD Getty support, as > may be > > desired. It seems to run fine with BSDGetty in my limited testing on a > > FreeBSD 10 installation on a virtual machine using an iaxmodem. > > > > http://sourceforge.net/p/hylafax/HylaFAX+/2339/ > > > > Daren, since you seem to have a working solution it may not matter to > > you to mess around with any more testing. You would have some limited > > benefit in updating to HylaFAX+ code, but it may not be worth the > effort > > to you. If you can easily mimick the build and installation from > > FreeBSD ports, then updating should be no problem. But it seems that > > this may be a bit more struggle for you, and you may be happiest to > just > > remain with your working solution. I'll send you a tarball of the code > > separately - saving you the effort of patching. If you don't want to > > use it, that's understandable and fine. > > > > If you (or anyone else) want to build and install HylaFAX+ on > FreeBSD it > > is this easy: > > > > ./configure -with-LIBTIFF="-L/usr/local/lib -ltiff" > > -with-TIFFINC="-I/usr/local/include" && make && make install > > > > If you want to use SysV Getty instead of BSD Getty, then change the > > configure command to: > > > > ./configure -with-LIBTIFF="-L/usr/local/lib -ltiff" > > -with-TIFFINC="-I/usr/local/include" -with-GETTY=SysV > > > > It may be more-involved if you want to get things going like JBIG and > > color fax support, etc. Someone interested should really volunteer to > > be the FreeBSD HylaFAX ports maintainer. > > > > Thanks, > > > > Lee.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?53BFE47D.4030203>