.comment: Road Building
Averting a Buffer Overflow

Dennis E. Powell
Wednesday, May 30, 2001 10:43:22 AM
I was able to get kdelibs started at
5:08 a.m., and scrolled to "make" at 5:12, when I turned
south on I-684. The compile continued as I turned off on the Sawmill
River Parkway, as I turned onto the Sprain Brook Parkway, as I
crossed the George Washington Bridge, as I headed down the New Jersey
Turnpike, as I paid the toll at the Delaware (state motto:
"Interesting corporate law and a state budget financed entirely
by traffic tickets") Memorial Bridge. It was welcomed by
Maryland, and though the back up on the Washington D.C. Beltway due
to a traffic accident seven hours earlier caused me to spend an
unscheduled two hours parked on that roadway, the compile of kdelibs
continued unhindered. It finally got done just as I was pulling into
a filling (and emptying) station in Springfield, Virginia, where I
needed to fill the gas tank and do what I could to minimize permanent
kidney damage from too much coffee combined with an unanticipated
Beltway delay. (The rental car company will never know how happy it
is that the traffic did finally start moving.)
So it was that just before noon I
started the compile of kdebase. It cooked along through Virginia and
most of North Carolina before it errored out. The problem was
something with the theme manager.
Let me pause for a moment to suggest
that there's been all together too much made of themes, in KDE and
everywhere else. They're fine, but they're certainly not essential.
They and their manager belong not in kdebase, which is not optional,
but in kdetoys, which is. Their current placement only gives
ammunition to those who argue that we're not quite grownups.
Anyway, I crossed my fingers, hit the
up arrow, added an "-i" to "make," and hit enter.
This seemed to satisfy kdebase, because it compiled its way through
the rest of North Carolina, all of South Carolina and Georgia, and
well into Florida. By the time it finished, it was about 10 p.m. and
I had driven 1144 miles. I pulled over and got a room for the night
In Daytona Beach, home of fast cars and the second-worst motel I've
ever stayed in (the worst being in Pensacola in 1997).
It looked as if I'd be at my
destination before KDE got built, but the other packages are smaller
and compile more quickly. Bright and early next morning, I headed out
in heavy I-95 traffic, kdeadmin starting to do its stuff. It finished
about Jupiter (where a fellow pulled out in front of me, causing me
to hit the brakes hard and mutter at him, "you may be in a
Saturn, but your head is in Uranus"), and I started kdegames. It
finished a few minutes after I arrived in Fort Lauderdale, where I
was stopping for the day to see family.
Next morning I headed out, now with my
wife, who'd flown in the night before, aboard, ready for the several
hours to Key West. Just as we left the motel I started the build of
kdegraphics. It finished just as we were entering Key Largo. Next up
was kdemultimedia, which chugged along up to and including the time
spent having lunch at ChilliWillie's in Islamorada, whose mascot is a
penguin. I started kdenetwork, which was still compiling in the car
as we checked into the guest house in Key West.
In fact, there were 11 additional
hours of compiling before, finally, koffice (which isn't technically
part of the kde distribution) blew up somewhere in killustrator and
could not be made to continue.
So, then, the answer is that it's not
even close -- one could drive to Key West from Connecticut and nearly
halfway back in the time it takes to compile the KDE-2.2 beta on my
dinky notebook. Which proves -- what? Well, nothing, except that you
could easily build KDE-2.2 beta on a ride from, say, New York to Los
Angeles. And, of course, that truly herculean efforts in pursuit of
absolutely nothing useful can make an otherwise boring ride a little
more amusing.
Having built the beta, I realized that
I was in no position to check out the features that most interest me
-- the typeface handling (the notebook obstinately refuses to run any
version of XFree86 that supports anti-aliasing) and the new print
engine (I brought no printer). The theme engine, of course, doesn't
work -- big deal -- though I suspect that that's probably fixed in
the current code, given the disproportionate emphasis currently
placed on themes. The rest of the apps seem to work just fine, and
though I had no real problems with any of them in the alpha release
of a few weeks ago, there is a certain, indefinable feeling of
robustness in the beta that was absent in the alpha.
I'll look at it a lot more closely
when I get home -- this is written from a motel room in Florida;
after a few hours' sleep I'll head back north. And now, having some
fairly accurate times for compilation on the notebook, I hope to
compare them with the same work done on a big, fast, desktop with
scads of memory. It will be interesting, along the way, to try to
learn whether chip speed or amount of memory is the more limiting
factor; I suspect that there are curves in both cases that intersect
at both top and bottom. But that's for another day. Now to find a
mirror that will give me 56k on this motel phone line. Wonder how
many miles it takes to compile Open Office . . .
« Back: Compiling on the Fly