|
Hylafax Developers Mailing List Archives
|
[Date Prev][Date Next][Thread Prev][Thread Next]
[Date Index]
[Thread Index]
[hylafax-devel] Re: missing parts of security-patch.sh
--------
Your message dated: Sun, 18 Jun 2000 17:17:06 +1000
>At 05:46 PM 6/17/00 -0700, Andy Sparrow wrote:
>>Namely:
>>
>>i) Testing shell variables that may not be set is dangerous,
>> e.g.:
>>
>> if [ -f $SYSINIT ]
>> ...
>>
>> is dodgy if 'SYSVINIT' is never set.
>
>SYSVINIT should *always* be set.
Then using:
if [ -f ${SYSVINIT:?} ]
guarantees it, if it's an error for it not to be set at that point.
if [ -f "$SYSVINIT" ]
may make the script fail more gracefully (at the risk/cost of hiding
the error until later).
As a matter of style, I invariably use double quotes around the
variable in the test. I often use the word substitution/checking form
too, but then I tend to structure my scripts so that the first usage
of them is extremely easy to locate (e.g. non-conditional).
This buys you robustness, in that you can screw around with the top of
the script/configuration/variable setup later, and it doesn't break
silently in the third included script of a 16,000 line installation
suite someone else wrote 4 years ago, running in a sub-shell in the
background where debugging it is a bitch...
> Though looking at configure it seems this
>has somehow being mucked up....and the Makefile in the etc directory seems
>broken now too.
I noticed it when the configure script started giving me 'start:
Command not found' messages. It wasn't hard to spot :-)
Hard part (for me, anyway), was working out how it ran that line of
code, given that I "knew" how /bin/test works.... :-/
Then I peered at the code for an hour, trying to spot the bug, before
re-reading the man page (carefully, this time) and realising it was
/implemented/ that way on purpose...
Ahhh, you stop learning, you start dying, ain't that the truth...
>> Of course, on BSD-ish systems, 'SYSINIT' doesn't get set...
Drat, wish I could type SYSVINIT consistantly...
>
>On BSD type systems the sysvinit script should be copied to the sbin
>directory instead.
>
>So that you can start and stop hylafax easily at the command line ie
>'hylafax start' and 'hylafax stop'
Yeh, the port was taking care of that by copying a 'hylafax.sh.sample'
file to ${PREFIX}/etc/rc.d (PREFIX being set to /usr/local).
This is a FreeBSD-ism. The port tells you that you have to take
manual, post-install action to enable automatic startup on re-boot.
This is deliberate.
Semantics for start-up scripts vary on BSD-ish systems though.. It's
not required to provide an argument (default is to start), although
it's not prohibited to supply one either.
I don't wanna revisit that particular TFH either (I actually /like/
the post-SysVr2 init.d directory, although I /still/ think it's kinda
overkill and ugly for a small machine).
>>ii) Some systems break the Adobe Font Metrics out of the
>> Ghostscript package altogether and install it as a separate
>> dependancy (presumably so that multiple ghostscript installs
>> can be located in versioned directory names, in a similar
>> way to how I can have multiple versions of Tcl/Tk installed
>> on the same system).
>
>Are you saying the metric files are separate from the fonts themselves?
Yes indeedy. See a couple of annotations down.
>> On such systems (FreeBSD at least, maybe also the other BSD
>> systems that use the 'ports' system), PATH_AFM would get set
>> just fine by the "last ditch" code if it weren't for the call
>> to the 'getGSFonts' function - which obligingly(?) finds the
>> Ghostscript 'afm' directory instead, thus no further
>> processing is performed.
>
>Er i dont understand what you mean - the idea is that the document is
>generated using the *same* fonts that ghostscript is going to use to
>actually rasterize the document.
Well, on re-reading, I wasn't very clear...
See next annonation for (hopefully) a better description.
>> However, this particular 'afm' directory doesn't have the
>> Courier font definition required for the package to run,
>> thus nothing works when the build is complete.
>
>
>Does it have other .afm files with encoded names? If so textfmt et al will
>use the Fontmap file to find the correct font.
Ooops, my bad. I reported this incorrectly (hey, it /was/ last week).
Reading through 'faxsetup.sh' again, I now recall that it blew up
saying it couldn't find Courier-Bold in the directory 'configure'
selected for it (/usr/local/share/ghostscript/fonts).
It selects the ghostscript 'fonts' directory because this code
(starting @ #3594 of 'configure' in 4.1b2):
if [ -z "$FONTMAP" ]; then
FONTMAP=`getGSFonts`
fi
Note "Setting the Fontmap path to $FONTMAP"
#
# Location of Adobe Font Metric files.
#
if [ -z "$PATH_AFM" ]; then
# if the fontmap path is available use that
# else just guess
if [ -n "$FONTMAP" ]; then
PATH_AFM=$FONTMAP
else
DIR_AFMS="
/usr/lib/afm
/usr/local/lib/afm
/usr/local/share/ghostscript/fonts
/usr/local/lib/ghostscript/fonts
/usr/share/ghostscript/fonts
/usr/gnu/lib/ghostscript/fonts
/opt/gnu/lib/ghostscript/fonts
/usr/lib/ghostscript/fonts
/usr/lib/gnu/ghostscript/fonts
"
case $TARGET in
*-irix*) DIR_AFMS="/usr/lib/DPS/AFM $DIR_AFMS";;
*-bsdi*) DIR_AFMS="/usr/contrib/lib/flexfax/afm $DIR_AFMS";;
*-sunos*) DIR_AFMS="/usr/openwin/lib/fonts/afm $DIR_AFMS";;
esac
PATH_AFM=
for i in $DIR_AFMS; do
test -d $i && { PATH_AFM=$i; break; }
done
fi
fi
if [ -z "$PATH_AFM" ]; then
# put it where ghostscript normally puts things
PATH_AFM=/usr/local/lib/ghostscript/fonts
Note "WARNING, could not locate a directory with font metric information,"
Note "guessing that font metric information goes in $PATH_AFM."
else
Note "Looks like font metric information goes in $PATH_AFM."
fi
(on FreeBSD at least), checks for FONTMAP, which is unset. Thus it calls
'getGSFonts()', and sets FONTMAP to the location of
'/usr/local/share/ghostscript/fonts' - which on this system, /doesn't/
contain the AFM files, as they're installed separately, as a
dependancy of /all/ of the Ghostscript versions.
Thus, in the next test, PATH_AFM is checked, found empty, and, as
FONTMAP is non-empty, PATH_AFM is set to FONTMAP. If it failed that
test, it'd run what I coined the "last-ditch" code, which would search
a list of directories for the AFM files, and it'd be set correctly,
because the correct location (for FreeBSD systems) is /before/
/usr/local/share/ghostscript/fonts in that list.
My hack was simply to avoid running 'getGSFonts()' if the system was
FreeBSD,and set the FONTMAP to the "correct" location. (And,
incidentally, I fat-fingered the patch - the call to 'Note' should
appear either outside the 'case' statement or in both clauses of it,
ideally).
To address your other question, here's the contents of the 'afm' and
'fonts' directories as installed:
pc# pwd
/usr/local/lib/afm
pc# ls
AvantGarde-Book Helvetica-Oblique
AvantGarde-BookOblique NewCenturySchlbk-Bold
AvantGarde-Demi NewCenturySchlbk-BoldItalic
AvantGarde-DemiOblique NewCenturySchlbk-Italic
Bookman-Demi NewCenturySchlbk-Roman
Bookman-DemiItalic Palatino-Bold
Bookman-Light Palatino-BoldItalic
Bookman-LightItalic Palatino-Italic
Courier Palatino-Oblique
Courier-Bold Palatino-Roman
Courier-BoldOblique Symbol
Courier-Oblique Times-Bold
Helvetica Times-BoldItalic
Helvetica-Bold Times-Italic
Helvetica-BoldOblique Times-Oblique
Helvetica-Narrow Times-Roman
Helvetica-Narrow-Bold ZapfChancery-MediumItalic
Helvetica-Narrow-BoldOblique ZapfDingbats
Helvetica-Narrow-Oblique
pc# file *
AvantGarde-Book: ASCII font metrics
AvantGarde-BookOblique: ASCII font metrics
AvantGarde-Demi: ASCII font metrics
AvantGarde-DemiOblique: ASCII font metrics
Bookman-Demi: ASCII font metrics
Bookman-DemiItalic: ASCII font metrics
Bookman-Light: ASCII font metrics
Bookman-LightItalic: ASCII font metrics
Courier: ASCII font metrics
Courier-Bold: ASCII font metrics
Courier-BoldOblique: ASCII font metrics
Courier-Oblique: ASCII font metrics
Helvetica: ASCII font metrics
Helvetica-Bold: ASCII font metrics
Helvetica-BoldOblique: ASCII font metrics
Helvetica-Narrow: ASCII font metrics
Helvetica-Narrow-Bold: ASCII font metrics
Helvetica-Narrow-BoldOblique: ASCII font metrics
Helvetica-Narrow-Oblique: ASCII font metrics
Helvetica-Oblique: ASCII font metrics
NewCenturySchlbk-Bold: ASCII font metrics
NewCenturySchlbk-BoldItalic: ASCII font metrics
NewCenturySchlbk-Italic: ASCII font metrics
NewCenturySchlbk-Roman: ASCII font metrics
Palatino-Bold: ASCII font metrics
Palatino-BoldItalic: ASCII font metrics
Palatino-Italic: ASCII font metrics
Palatino-Oblique: ASCII font metrics
Palatino-Roman: ASCII font metrics
Symbol: ASCII font metrics
Times-Bold: ASCII font metrics
Times-BoldItalic: ASCII font metrics
Times-Italic: ASCII font metrics
Times-Oblique: ASCII font metrics
Times-Roman: ASCII font metrics
ZapfChancery-MediumItalic: ASCII font metrics
ZapfDingbats: ASCII font metrics
pc# cd /usr/local/share/ghostscripts/fonts; ls
a010013l.afm c059013l.pfm hritro.gsf n019024l.pfb n022023l.pfm
a010013l.pfb c059016l.afm hrpld.pfa n019024l.pfm n022024l.afm
a010013l.pfm c059016l.pfb hrpldb.gsf n019043l.afm n022024l.pfb
a010015l.afm c059016l.pfm hrpldbi.gsf n019043l.pfb n022024l.pfm
a010015l.pfb c059033l.afm hrpldi.pfa n019043l.pfm p052003l.afm
a010015l.pfm c059033l.pfb hrplr.gsf n019044l.afm p052003l.pfb
a010033l.afm c059033l.pfm hrplrb.gsf n019044l.pfb p052003l.pfm
a010033l.pfb c059036l.afm hrplrbo.gsf n019044l.pfm p052004l.afm
a010033l.pfm c059036l.pfb hrplro.gsf n019063l.afm p052004l.pfb
a010035l.afm c059036l.pfm hrpls.gsf n019063l.pfb p052004l.pfm
a010035l.pfb d050000l.afm hrplsb.gsf n019063l.pfm p052023l.afm
a010035l.pfm d050000l.pfb hrplsbo.gsf n019064l.afm p052023l.pfb
b018012l.afm d050000l.pfm hrplso.gsf n019064l.pfb p052023l.pfm
b018012l.pfb fcyr.afm hrplt.pfa n019064l.pfm p052024l.afm
b018012l.pfm fcyr.gsf hrpltb.gsf n021003l.afm p052024l.pfb
b018015l.afm fcyri.afm hrpltbi.gsf n021003l.pfb p052024l.pfm
b018015l.pfb fcyri.gsf hrplti.pfa n021003l.pfm putb.pfa
b018015l.pfm fhirw.gsf hrscc.pfa n021004l.afm putbi.pfa
b018032l.afm fhirw.pfm hrsccb.gsf n021004l.pfb putr.pfa
b018032l.pfb fkarw.gsf hrscco.gsf n021004l.pfm putri.pfa
b018032l.pfm fkarw.pfm hrscs.pfa n021023l.afm s050000l.afm
b018035l.afm fonts.dir hrscsb.gsf n021023l.pfb s050000l.pfb
b018035l.pfb fonts.scale hrscso.gsf n021023l.pfm s050000l.pfm
b018035l.pfm hrger.pfa hrsyr.gsf n021024l.afm u003043t.afm
bchb.afm hrgerb.gsf n019003l.afm n021024l.pfb u003043t.gsf
bchb.pfa hrgerd.gsf n019003l.pfb n021024l.pfm u003043t.pfm
bchbi.afm hrgero.gsf n019003l.pfm n022003l.afm u004006t.afm
bchbi.pfa hrgkc.gsf n019004l.afm n022003l.pfb u004006t.gsf
bchr.afm hrgks.gsf n019004l.pfb n022003l.pfm u004006t.pfm
bchr.pfa hrgrr.pfa n019004l.pfm n022004l.afm z003034l.afm
bchri.afm hrgrrb.gsf n019023l.afm n022004l.pfb z003034l.pfb
bchri.pfa hrgrro.gsf n019023l.pfb n022004l.pfm z003034l.pfm
c059013l.afm hritr.pfa n019023l.pfm n022023l.afm
c059013l.pfb hritrb.gsf n019024l.afm n022023l.pfb
where the types are (e.g.):
pc# file a*
a010013l.afm: ASCII font metrics
a010013l.pfb: PostScript Type 1 font program data
a010013l.pfm: raw G3 data, byte-padded
etc.
Incidentally, 'fonts.(dir|scale)' don't contain any definitions for
'Courier'.
>>iii) You might want to add these lines to 'configure' around
>> about the definition settings for the socket API (roughly
>> line 2311 in my copy):
>>
>> *-freebsd4*) CONFIG_SOCKARGLENTYPE=u_int32_t;;
>> *-freebsd5*) CONFIG_SOCKARGLENTYPE=u_int32_t;;
>
>...are you sure the above isn't the 'size_t' type, that seems to be what
>alot of the other platforms are using.
Ummm, Yes, quite sure, see the definitions below...
FreeBSD 4.0-STABLE (and 5.0-CURRENT) is the chosen development platform
of the KAME Project (integrated IPv6 stack and tools & IPSec for IPv4/v6,
see http://www.kame.net).
The KAME code is an integrated part of FreeBSD since 4.0-RELEASE.
OpenBSD 2.7 and up have it too, add NetBSD if you're talking development
trees.
Perhaps that's the reason for the changes.
FreeBSD 3.4-STABLE #3: Sun Apr 2 21:22:34 PDT 2000:
* @(#)socket.h 8.4 (Berkeley) 2/21/94
* $FreeBSD: src/sys/sys/socket.h,v 1.27.2.4 2000/03/08 06:32:44 fenner Exp $
*/
#ifndef _SYS_SOCKET_H_
#define _SYS_SOCKET_H_
/*
* Definitions related to sockets: types, address families, options.
*/
/*
* Data types.
*/
typedef u_char sa_family_t;
typedef int socklen_t;
FreeBSD 4.0-STABLE #0: Mon May 22 21:32:32 PDT 2000:
* @(#)socket.h 8.4 (Berkeley) 2/21/94
* $FreeBSD: src/sys/sys/socket.h,v 1.39 2000/03/11 19:51:04 shin Exp $
*/
#ifndef _SYS_SOCKET_H_
#define _SYS_SOCKET_H_
/*
* Definitions related to sockets: types, address families, options.
*/
/*
* Data types.
*/
typedef u_char sa_family_t;
typedef u_int32_t socklen_t;
HTH
Regards,
AS
____________________ HylaFAX(tm) Developers Mailing List ____________________
To unsub: mail -s unsubscribe hylafax-devel-request@hylafax.org < /dev/null
