62 lines
3.1 KiB
Plaintext
62 lines
3.1 KiB
Plaintext
|
Wed Sep 15 00:29:10 EDT 1999
|
||
|
|
I tuned ifacechk to be able to handle _huge_ numbers of packets. Perl was
|
||
|
|
maxin' out at some point ... who cares. I fixed it with line #113:
|
||
|
|
s/([0-9]+)([0-9]{9})/$2/g := just gimme the last 9 digits... work on that.
|
||
|
|
This forces a rollover on 9 digits, simultaneously causing correct operation
|
||
|
|
999999999 out of 10000000000 times (assuming a 1 byte per second xfer rate and
|
||
|
|
an update delay of 1 second). It also causes an error on the rollover (fixed
|
||
|
|
after the next update. This is not an elegant solution, but, it is
|
||
|
|
satisfactory for now.
|
||
|
|
|
||
|
|
Tue Aug 24 12:14:38 EDT 1999
|
||
|
|
I just noticed a bug in ifacechk. The code that reverses the logical/physical
|
||
|
|
devices doesn't work quite right. Apparently if /var/run/ppp-ppp0.dev
|
||
|
|
contains 'ppp1', then it means that the physical device 'ppp0' is acting as
|
||
|
|
the logical device 'ppp1'. That's news to me. The reversal was working
|
||
|
|
before, because if both devices are up, the reversal is correct even counting
|
||
|
|
my error. Did you know that Keppler made two glaring errors in his calculus,
|
||
|
|
and the second cancled out the first? I guess I'm up there with greatness. ;)
|
||
|
|
|
||
|
|
Tue Aug 24 12:07:07 EDT 1999
|
||
|
|
I added some better error checking. When one of my devices would go down,
|
||
|
|
wmjiface would start showing me non-sense while the device is coming back
|
||
|
|
up. If an error is detected, the exponential averaging get's cleared out.
|
||
|
|
That helps to keep it from holding on to negative speeds for long periods.
|
||
|
|
|
||
|
|
Wed Aug 18 02:22:43 EDT 1999
|
||
|
|
I finnally installed the weighted averaging I was talking about! exp_avg =
|
||
|
|
(alpha*current_bps) + ((1-alpha)*last_average). This method is called
|
||
|
|
exponetial averaging. I had to install a "beta pulldown," where alpha is
|
||
|
|
equals beta iff the xfer_rate exactly equals 0. It gives a more realistic
|
||
|
|
feel to the end of the xfer, rather than a drawn out approach to 0. That
|
||
|
|
limit could go on for tens of minutes, since I used doubles (so the history
|
||
|
|
will contain a more accurate approximation).
|
||
|
|
|
||
|
|
Tue Aug 17 22:48:18 EDT 1999
|
||
|
|
Made a minor change that reverses my ppp devices when needed. The ppp devices
|
||
|
|
are fcfs btw, I don't want the imaginary device, I want the real device.
|
||
|
|
|
||
|
|
Tue Aug 17 03:47:06 EDT 1999
|
||
|
|
Somewhat drunkenly completed a change to ifacechk to remove the old tmp files.
|
||
|
|
Ended up using a system() if ... /proc/dir exists check ... how ugly.
|
||
|
|
but hey... I was already using /proc *smirk*
|
||
|
|
|
||
|
|
Mon Aug 16 22:07:05 EDT 1999
|
||
|
|
noticed an ugly flaw in ifacechk. Every instance of ifacechk was overwriting
|
||
|
|
every other's data ... I used perl's '$$' variable to seperate it out. The
|
||
|
|
only trouble is that there will now be a huge buildup of the stat files after
|
||
|
|
a while. How to resolve that?
|
||
|
|
|
||
|
|
Mon Aug 16 22:06:15 EDT 1999
|
||
|
|
added a new switch (and the support for it). -D: now lists in/out at the
|
||
|
|
same time.
|
||
|
|
|
||
|
|
Mon Aug 16 20:06:18 EDT 1999
|
||
|
|
huge changes to the update loop in the form of more modularization.
|
||
|
|
|
||
|
|
Tue Jul 20 11:08:38 EDT 1999
|
||
|
|
installed new date and version stuff in the makefile, and the print_help()
|
||
|
|
|
||
|
|
Wed Jul 14 19:08:20 EDT 1999
|
||
|
|
It's new... There's no changes... besides it's existance.
|