Re: Install of HylaFax on RedHat Linux release 4.1

["David A. De Graaf" <degraaf@digitel.net>]

Johannes Joubert wrote:
> 
> I would like to install a new version of Hylafax on RedHat Linux release
> 4.1 (kernel 2.0.27)
> 
> If there is anybody out there who did such an installation could he please
> tell me what packages he used and how he set it up?
> 

Sure.  I've just accomplished that.
In the process I rediscovered   :-(   all the things that Matthew Rice
tells of in
ftp://ftp.sgi.com/sgi/fax/binary/INSTALL-hylafax-i386-linux-v4.0pl1

To his excellent notes I would only add this amplification.

RedHat 4.0 Linux (kernel 2.0.18) includes all you need to build
tiff-v3.4beta035 and hylafax-v4.0pl1.
	gcc 2.7.2
	ld 2.7
	gnu make 3.74
	gnu ar 2.7
	gnu awk 3.0 pl0
	gnu sed 2.05
	  (what else???)
You should probably replace ghostscript 3.33 with the better ver 4.0.

Although the configure script for HylaFAX performs miracles, be aware
of these problem areas:
1)  If you want the html files installed you must edit configure to set
	HTML=yes
If you have ready access to the Internet this is a little silly.
http://www.vix.com/hylafax is a better place to look.

2)  Configure looks in many places for the ghostscript AFM fonts, all
of them wrong.  You can either edit configure to add the right place
or correct its wrong deduction in the interactive phase.
Mine are in /usr/local/share/ghostscript/fonts.

3)  Similarly, configure will incorrectly deduce where the lock
files are.  You should run each of the out-calling programs, cu,
uucico, kermit, etc. and observe where they put them, so you can
correct configure's wrong answer.  I found them in /var/lock.
If you allow hylafax to compile with the incorrect lock file location,
when faxgetty runs, cu, et al, will connect to the port but be unable
to receive data from the modem.  All characters are gobbled up by
faxgetty instead, because it can't see the lock file.

4)  According to /usr/doc/getty_ps-2.0.7h-3/README.linux,
/dev/ttyS# is used for call-in (getty), and /dev/cua# for call-out
(cu, uucico, kermit, etc.)  In addition, my system came with a
symbolic link,  /dev/modem -> /dev/cua1, and the ppp script used that.

However, HylaFAX is designed to use just one device port for both
incoming and outgoing data and fax calls, and it should be the one
normally used for out-calling - /dev/cua1.  I think it is desirable, if
not mandatory, for all the other communication programs to use that
same port, so the lock files will be consistently named.


BTW, since HylaFAX is such an "interesting" package, it invites
benchmarking.  Instead of merely doing
	make
try this:
	time make 2>&1 | tee make.out

Here are some elapsed times I've observed:
HW		SW		elapsed time mm:ss
Pentium 200	Linux RH4.0	7:28 (usr 270.85s, sys 69.07s)
Pentium 133	Solaris 2.5	9:26
486/33		Dell UNIX	53:

-- 
	Dave De Graaf,  Internet: degraaf@digitel.net
	DATIX, Inc.	voice: (803) 785-3136	fax: (803) 785-3156