[lug] File modification
mc at morat.net
Thu Mar 14 06:05:34 MST 2002
-----BEGIN PGP SIGNED MESSAGE-----
On Wednesday 13 Mar 2002 16:45, you wrote:
> From my experience, working in "the industry" will be just like trying
> to get class assignments done and working the night before they're due.
And the assignments will be given out after they're due.
> There _always_ seems to be some last minute items that wasn't accounted
> for in the schedule. And it's always hard to schedule the unexpected
> problems that seem to appear.
I've just spent about a week and a half trying to get a PDF program working.
It worked fine for me - but then we create it as a shared lib (it's C++) and
call it through JNI from an RMI interface. That part doesn't work so well.
Turns out it leaks memory like a sieve. Ok, I go back and find a couple of
stupid mistakes in my code, fix them, it still leaks. The provider of the
library (not open source, not my call) says they have a lot of people using
this library without any problems. After 4 or 5 days of this they volunteer
to take a look at my code (mainly, I think, to prove the problem is in my
code :) and I get a reply back the next day: "Oh, you were supposed to call
this function before the object goes out of scope - it turns out the
destructor doesn't clean up all the memory. We're fixing that."
Ok, that's one problem solved - now it still doesn't work under the multiple
threading environment. Note that my code isn't internally threaded, it just
has to run in multiple threads. Again, a couple of things on my side that
weren't thread safe, fix them, it still doesn't work. Back to the company. I
get a "more thread safe" version of the library. It's better, but it still
doesn't work. *sigh* After a couple more days of discussion I get a version
of the library that was compiled with -fPIC. Initial test looks good -
process 100,000 documents in a bit over 2 hours.
Package it up in the form to be called from java, send it off, start a new
test. So far I've processed between 600,000 and 700,000 documents (test
program runs 100 threads, keeps cycling between the 4500 input documents) in
about 16 hours. Looks like we have a winner.
In spite of the problems with the library I would recommend these folks if
you happen to need a commercial PDF library. The documentation is a bit
lacking (personal opinion, it might be my lack of familiarity with the
C++/PDF combo that influences that opinion :) but they were quick to help and
answer questions and such.
Ken Weinert mc at morat.net
GnuPG KeyID: 9274F1CE GnuPG available at http://www.gnupg.org/
GnuPG Key Fingerprint: 1D87 3720 BB77 4489 A928 79D6 F8EC DD76 9274 F1CE
Copywight 1994 Elmer Fudd. All wights wesewved.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org
-----END PGP SIGNATURE-----
More information about the LUG