PC-BSD

A look at a *good* PBI install script from 2007.

by

In looking around my ~/Projects folder, I found an old PBI install script I wrote a couple years back. When I was working on a TexLive PBI (only me and Oko were interested), I wrote a very robust script. Most PBI that use install scripts should be nearly like this but normally, they only handle the GUI case and screw you if you use text mode or have special need.

This script was written to the published API at the time, which basically amounted to $INSTALLMODE being set to tell you if you were in a GUI or not; a variable to tell you what name your PBI would have inside /Programs. This was an improvement over using “$1” everywhere, as was required in previous versions of the PBI API.

Here is my old script:

#!/bin/sh

# the year of this release, e.g. 2007, 2008 e.t.c. -> This should be equal to
# the Program Version we set in PBC.
YEAR="2007"

# the size of all our installed files in $PBI_BASE/texlive/$YEAR/
TEXSIZE="1.1GB" # as a string

# compat check.. should be unecessary but I don't trust PC-BSD. Since users can
# now override our install location and documentation does not talk to us about
# coping with it not being /Programs/$PROGDIR/, abort install rather then risk
# performing undefined/undocumented behavior should /Programs/$PROGDIR be
# overrided.

if [ -d "/usr/Programs" ]; then
    PBI_BASE="/usr/Programs"
elif [ -d "/Programs" ]; then
    PBI_BASE="/Programs"
elif [ -d "/usr/local/MyPrograms" ]; then
    PBI_BASE="/usr/local/MyPrograms"
else
    if [ $INSTALLMODE = "GUI" ]; then
        kdialog --sorry "Can't find PBI Installation Directory... aborting"
    else
        echo "Can't find PBI Installation Directory... aborting"
    fi
    exit 255
fi

# set the path
MY_TEX_PATH="$PBI_BASE/texlive/$YEAR/bin/i386-freebsd/"


# XXX check if we are already installed or improperly installed

if [ -d "$PBI_BASE/texlive/$YEAR" ]; then
    if [ $INSTALLMODE = "GUI" ]; then
        kdialog --sorry "$PROGDIR appears to be already installed, aborting"
    else
        echo "$PROGDIR appears to be already installed, aborting"
    fi
    exit 255
fi

# give the user a chance to abort the installation
if [ $INSTALLMODE = "GUI" ]; then 
    kdialog --warningyesno 
    "The installation requires approximatly $TEXSIZE of disk space Continue?"
    if [ $? -gt 0 ]; then
        exit 1
    fi
else
    echo "The installation requires approximatly $TEXSIZE of disk space"
    echo -n "Continue?[yes/no]"; read GO
    echo $GO | grep -i "no"
    if [ $? = 0 ]; then
        exit 1
    fi
fi


# Do installation
echo 'MSG: Setting up TeX Live directory'
mkdir -p $PBI_BASE/texlive/$YEAR
ln -s $PBI_BASE/texlive/$YEAR $PBI_BASE/$PROGDIR/TEXLIVE_$YEAR

echo "MSG: Installing TeX Live files..."

# extract our texlive installation
cd /$PBI_BASE/texlive/$YEAR/ && 
    lzma d texlive${YEAR}.tar.lzma -so | tar -xpf -

# prompt for a default paper size
if [ $INSTALLMODE = "GUI" ]; then 
    PAPER=`kdialog --combobox "Select default paper size" 
            "A4 Paper" "US Letter"`
else
    echo -n "default paper size [A4 Paper, US Letter]: "; read PAPER
fi

echo $PAPER | grep -i "Letter"
if [ $? -eq 0 ]; then
    texconfig-sys paper letter
fi

echo $PAPER | grep -i "A4"
if [ $? -eq 0 ]; then
    texconfig-sys paper a4
fi


echo "MSG: Updating TeX filename databases"
texconfig-sys rehash


if [ $INSTALLMODE = "GUI" ]; then 
    kdialog --yesno 
    "Congratulations! Installation complete -- test installation?"
    if [ $? = 0 ];then
        cd /tmp/
        $MY_TEX_PATH/pdflatex sample2e.tex && kpdf /tmp/sample2e.pdf
        if [ $? -gt 0 ]; then
            kdialog --error 
        "The test may have failed to load, please report any errors: $?"
        fi
    fi
    echo "MSG: Displaying readme file"
    kwrite $PBI_BASE/$PROGDIR/PBI_README.txt 
else
    # I don't know a programa off hand to use for console output.
    more $PBI_BASE/$PROGDIR/PBI_README.txt
fi

What is wrong with that script? Nothing compared to what is wrong with the PBI script API.

$PROGDIR should expand to /Programs/ProgNameVer instead of ProgNameVer.

The maintainer shouldn’t have to treat GUI and TEXT mode differently – the API should do it. In fact, that was the main impetus for the EPI spec creating platform independence. In an EPI, people would have said ask the user a question. Not have to understand shell script and massive toggles for GUI/TEXT mode install.

I believe PBI now also do similiar idiocy in regards to fonts handling that should be done at a higher level. However, according to the officially published documentation, my old script is still the correct way to go with a PBI.

One of the reasons I gave up on Kris Moores PBI, was the brain damaged design, the other reasons were the total miss management of PBI…

Under the draft spec’s I wrote for EPI, the above script would have become something like this:

# optional EPIL script for EPI

define question dialog with
        title = "TextLive EPI"
        message = "Please set a default paper size"
        default = "A4 Paper"
        options = ['A4 Paper', 'US Letter']
launch as paper_size

if paper_size == "A4 Paper" then do
        execute "texconfig-sys paper a4"
else
        execute "texconfig-sys paper letter

Now I ask you, if you were not a programmer, which one do you understand? Then let me ask you, if you just want to get on with your life, which would you rather write?

Both the PBI sh and EPI EPIL scripts would result in a simple dialog asking you to choose from one of two options. If you know how to use TexLive, the EPIL script would be totally unnecessary because you already know how to change paper sizes. That means you don’t even have to write a script at all in a normal case.

End result? Oh so simpler.

A few Thoughts turn to PC-BSD

by

Considering recent threads in the PC-BSD forums, largely centred around the projects deficiencies I’ve made a decision.

I have the DVD version of PC-BSD 7.1.1 x86 down loading, when it is complete, I’ll load it into a virtual machine for testing. Once that is done, I will conduct a ‘deep’ inspection of the systems current state—including the source code. The aim of it is to analyze the distributions weak spots, noting the areas where it does need improvement. After that is completed, I intend to publish my catalogue of the projects deficiencies along with suggested paths for improvement. It’ll probably be posted here on my LiveJournal, with links shared in the appropriate forums and mailing lists.

Why do I feel like doing this? Well, PC-BSD while it has some good qualities, generally fails (in my honest opinion) because of how it is handled. As I have often said and people can feel free to quote me on it, PC-BSD feels like a project born in Kris Moore’s garage: that never left. And like wise, as I have often expressed, PC-BSD development practices do *NOT* reflect those of a real BSD project.

In my experience getting anything done there, that is also worth doing: has usually been caused by the community screaming the developers ears off… perhaps consistently so. So writing an open letter might poke the project into cleaning up their ship, or at least demonstrate to the rest of us what “Not to do”, lol.

nights log

by

So far it has been a fairly uneventful night. I’ve been contemplating taking up another language, I’m interested in both Latin and Spanish, but have started studying neither yet. Mathematically, I would have to favour Spanish… because shall we say, while exposure to Spanish is fairly easy for me, Latin is a bit of a dead language! Really the only languages of value out here, is what *passes* for English, and of course Mexican / Puerto Rican dialects of Spanish. Considering the simpler grammar, studying Spanish in the long run: might also help to improve my German, which is useless at the compositional level, but occasionally handy at the reading level. I think the entire world would agree though, that I should work on improving my English skills, or at least stop the mixture of US/UK elements.

For time passing, I setup sudo on my laptop; had done it some years back, when I was running PC-BSD, but haven’t used it since an updated overwrote my sudoers file, as I had forgotten to backup /usr/local/etc. I knew that PC-BSD was taking ownership of /etc/rc.local and making overwriting changes to sudoers, but didn’t use sudo enough to care lol. That’s one thing that has always pissed me off about PC-BSD, yet I have always respected about the *real* BSD projects: a concept of user edited files. You can play with /etc/rc.local and /etc/rc.conf on FreeBSD without getting biten, on PC-BSD you should use an rc.conf.local file: OpenBSD style. I’m always running the latest FreeBSD STABLE branch on the laptop, in this case RELENG_7, and anxious to see RELENG_8 become a practical reality. Heh, even on the OpenBSD box (which as sudo preinstalled) I don’t bother using sudo that much. What I like about sudo however, is the ability to skip ‘su – root’, and limit access in a more general way—to specific commands. It’s probably one of the best programs of its kind written, since su.

In getting playful, I split up my word lists (:help spellfile) in vim, into smaller organized chunks, and added a few more entries from FreeBSDs system dictionaries. As time goes on, I might expand things further, based on the kinds of texts I end up editing. All of my text editing is basically done in vim, so it also does most of my spell checkin’ by virtue of that. Of course, Vi IMproved has an excellent spell checker, and perhaps the only one I’ve met that could be considered superior to MS Words. While I utterly hate word processors, and find them to be useless paper weights; I generally consider Microsoft Word the best known example of spell checking lol. I’m happy to say though, that I’ve evolved somewhat in the past eleven or twelve years ;). Now if only I could do something about the having-to-type-as-fast-as-i-can-freaking-think-problem, maybe my writing might actually be worth something, someday lol.

The ‘Burbs was on TV, and coming on in a little bit is a good film that I really haven’t seen in a long time. It’s called Throw Momma from the Train, best described as a comedic remake of Strangers on a Train. I plan to log off, lay down, and try to relax for a bit…. hopefully relax. Probably will try to dig up another quick snack, something light but useful.

The internet connection has been offline for the last couple hours, first found out it was up when an instant message popped up on me mid-sentence! Hmm, got about 10 minutes until the movie starts, so it’s time tos tart winding down. I’ll probably be awake early tomorrow, assuming I get any sleep before work….. sigh.

Pondering: should I ever go back (regularly)?

by

I used to spend a lot of time at the PC-BSD tech support forums (forums.pcbsd.org). As some should know, it is not one of my ‘favorite’ projects for numerous reasons.

I haven’t been active there in I guess, nearly a year now. Should we say, shuffling though like 50 threads a day and being one of the few paying attention got quiet old, when I’ve so much else to worry about after work. Ending up as little more then a spam-patrol man wasn’t very interesting either. Although things have been realitively docile since FreeBSD opened their own forum, I hang my hat over at DaemonForums.org – a place that’s worth visiting.

One of the alarming things about forums.pcbsd.org even while I am inactive, I still have several orders of magnitude more posts then everyone else… Man I used to check that forum several times a day. At one point, I may have been the last skilled snook left, I hope that has changed….

What I hate about programming

by

Some months ago when it reached Kris Moore’s attention (late as usual) that I had brought up security issues with his Firefox3 PBI, he changed it to something almost as bad. A couple weeks ago, I heard back from Kris that he had [naively] changed the code for making Fx3 the users default browser would no longer run as “root”. After a little more conversation he split it off to something better.

Originally it was a part of the script that runs during PBI installation (and worse then the below script), probably tired of my replies he made an extra wrapper around the Firefox3, that asks the user if they want Firefox3 set default or not, rather then workin’ the user database at install time. (I refuse comment on the following scripts predecessors: if you want to know more, read his SVN). The solution he came up for that wrapper, was to invokes the following code as the user when necessary:

#!/bin/sh
# Helper script to make FF the default browser for a user
##############################################################################

# Get the users homedir
USER="`whoami`"
HOMEDIR="`cat /etc/passwd | grep ^${USER}: | cut -d ":" -f 6`"

if [ -e "${HOMEDIR}/.kde4" ]
then
   KDEDIR=".kde4"
else
   KDEDIR=".kde"
fi

if [ ! -e "${HOMEDIR}/${KDEDIR}/share/config/kdeglobals" ]
then
  echo "ERROR: No kdeglobals file for $USER"
  exit 1
fi


TMPKFILE="${HOMEDIR}/.kdeglobals.$$"
TMPKFILE2="${HOMEDIR}/.kdeglobals2.$$"
rm ${TMPKFILE} >/dev/null 2>/dev/null

cat ${HOMEDIR}/${KDEDIR}/share/config/kdeglobals | grep -v '^BrowserApplication' > ${TMPKFILE}

rm ${TMPKFILE2} >/dev/null 2>/dev/null
touch ${TMPKFILE2}
while read line
do
  if [ "$line" = "[General]" ]
  then
    echo "$line" >> ${TMPKFILE2}
    if [ "${KDEDIR}" = ".kde4" ]
    then
      echo "BrowserApplication[$e]=!/Programs/bin/firefox3" >> ${TMPKFILE2}
    else
      echo "BrowserApplication=!/Programs/bin/firefox3" >> ${TMPKFILE2}
    fi
  else
    echo "$line" >> ${TMPKFILE2}
  fi
done < ${TMPKFILE}

# all finished, now move it back over kdeglobals
rm ${TMPKFILE}
mv ${TMPKFILE2} ${HOME}/${KDEDIR}/share/config/kdeglobals

exit 0

which is more secure then the original implementation, and more efficient also. Tonight I sent Kris a casual (read: adapt to need, don’t take as is) suggestion from yours truly: 

#!/bin/sh
# Helper script to make FF the default browser for a user
# Should work for KDE3 and KDE4.
##############################################################################

PROG="!/Programs/bin/firefox3"
FILE="./share/config/kdeglobals"

for D in "${HOME}/.kde" "${HOME}/.kde4"
do
   cd $D 2>/dev/null || break;

   if [ ! -e "$FILE" ]
   then
       echo "ERROR: No kdeglobals file, unable to set $PROG as default"
       exit 1
   fi

   ed -s "$FILE" <<EOF
       /[General]/
       /BrowserApplication.*=/
       s/=.*/=${PROG}/
       wq
   EOF
   # write your own error handlers
done

exit 0

which should work as far as I can test; since I lack a working KDE install (compiling KDE4.2+ is on my todo list). It’s not perfect, but it sure is nicer then what he had a few months back. I included the a diff of the two scripts in my last message, which may very well go against my decision to “never” send these people patches. But I really don’t care if he accepts it or not, because while I believe in being helpful, I also I do not like doing peoples jobs for them.

I’m a lazy good for nothing creep, but I am lazy of muscle – not lazy of mind. The most productive code I have ever written, is the code I was smart enough /not/ to write in the first place.

lol@someproject

by

I don’t know what is more funny, reporting potential security errors to a project or looking at the even worse ‘solutions’ they cook up.

Annoyance with half assed work

by

I will never understand why so many people seem to do so much insecure $h|+ with temporary files…. for the love of petes sister, if the system provides secure alterntiives — use them for cryoing out !)%(!%)!(%Y)!ing loud. Is it really that hard? It’s not like you have to implement the damn wheel everwhere…. And if osme one is going to reimplement the damn wheel — make it a better one, not even worse one shaped like a triangle!!!

Some times… people really, really, really annoy me with what they do….

I really hate to tell people how to do their job

by

I really hate to tell people how to do their job, especially when they are the ones who should be getting paid to do it; but sometimes I can’t help but wonder, if some people have ever heard of the words race condition or predictability, being used in the same sentence as exploit or security vulnerability?

Getting playful

by

I’ve been experimenting with extracting PC-BSD PBI on FreeBSD; mainly because I would like to do a little postmortem analysis on a couple of PBI’s, to see if the “Officially sanctioned” PBI Developers are still following the rules.

The process is actually a lot simpler then I thought it would be, the only problem is doing it in a chroot lol. Already more then 200MB in files in /tmp/chroot, and I’ve found that PBI must “assume” the presence of a lot of shit, because it will segfault if even simple things like awk or whoami are missing, let along bigger things 8=), what ever happened to error checking? But anyway, it’s a fairly easy thing to sort out, even with having to take a fair number of libraries and programs into the chroot, in order to fool the thing into -extract’ing. So far, I’ve only hit one snag:

/home/0/.PBItmp/.pbistart: ./PBI: not found

which I have not figured out yet.

I think I’ve also found the origin of one of the more stupid elements of the PBC Sh API, and it seems to reflect the PBI sub systems source, judging by what I see in PC-BSDs Subversion >_>. It also reminds me, that some people seem to have never heard of a symbolic constant in their entire lives…. which also explains a few other things about the PBI Creation process and PC-BSD in general. Since PC-BSD, unlike a *real* BSD system, does not believe in documenting anything. I’ve had to go straight to the (also undocumented) source for answers, it also temps me to write a detailed review and commentary – but I’ll keep my mouth shut for now. The number of people they will probably buttfuck in the long run is their own concern, I’ve already left. I’m not really interested in shifting through several thousand lines of intermixed Bourne, Bourne Again, and C++ code just to audit a few PBI; let some other poor schlep, eh Good Samaritan deal with this schlockware.

If my post doesn’t make it obvious by now, the reasons why I was conducting the tests within a chroot environment, rather then sparing myself the trouble -> I wouldn’t let the PC-BSD developers, or most peoples PBI touch one of my systems with a ten-thousand foot cattle prod, and I don’t have time for setting up a jail. That is the kind of feelings I’ve got for the project, after using PC-BSD for years; now OpenBSD, FreeBSD, and NetBSD on the other hand, them I trust… and have seen the resulting work to warrant it lol.

– A strict son of a bitch.

by

Well, it’s been on my list for awhile, but I finally got around to posting it on the PC-BSD forum. Poll: Common problem/questions FAQ for each release branch. It was one of the things on my list of project short comings, that reminds me… I need to add more >_>

One reason why I use FreeBSD and OpenBSD on my machines, they do a professional job of things lol. Maybe it’s the nature of projects like PC-BSD and Ubuntu, to pay less attention to things then the systems they’ve ripped use Both FreeBSD and Debian do a much better job them PC-BSD and Ubuntu at many things, hell even Slackware and Windows does better lol.

As to the thread I started on forums.pcbsd.org, I also sent out a PM to 20~30 people in the forum regular, semi-regular, and past regular visitor range, asking for their inputs. With a little luck, between anyone that actually reads their PMs and our current crowd, that’ll help contribute feedback in the thread, to prove I’m not just talking out my ass…