Spork Boards

Developers, Developers, Developers!

Cloudscout's Avatar Picture Cloudscout – February 25, 2008 12:31PM Reply Quote
Don't mind the pit-stains, greasy hair and Dorito breath. They've been busy coding. Or, at least, thinking about coding.

johnny k – April 05, 2011 12:20PM Reply Quote
If you're upfront with solutions, that will avoid them seeing you as part of the problem.

porruka (Admin) – April 05, 2011 12:23PM Reply Quote
Failure is pre-greatness.
Depending on your position in the hill-sh¡t hierarchy related to this particular closet, I might amend JW's excellent advice. You may be forced to triage the bones and, depending on the "ship within a week" requirement's absoluteness, present not only an accurate summary of current conditions, but what might be able to be fixed in that week relative to the PainNSuffering™ that will undoubtedly arise. But as JW says, without exaggeration; from the sounds of it, reality will be horrific enough on its own.

El Jeffe – April 05, 2011 01:22PM Reply Quote
What a journey.
A query for the more experienced developers.... if you're trying to tidy up a little skeleton left in a software closet, and it turns out that there's an entire ossuary stuffed back there, what do you do?

Can you be a bit more specific?

John Willoughby – April 05, 2011 01:59PM Reply Quote
Save us, Lord, from the furries of the Norsemen!
The predecessor in his position set up a software "dead man's switch" which, if he didn't log in for two months, uploaded all of the organization's proprietary information to Wiki Leaks and posted naked pictures of senior management to redtube.

Just guessing.

And, perhaps, subconsciously planning...

ARL (Moderator) – April 05, 2011 02:10PM Reply Quote
I whinge therefore I am!

You've been in YDD's shoes a few times I assume?

John Willoughby – April 05, 2011 02:43PM Reply Quote
Save us, Lord, from the furries of the Norsemen!
I've just noticed that when Bad Things happen to projects, Bad Things happen to engineers. I couple this observation with the one that goes, "You don't have to be faster than the bear, just faster than the slowest camper" to formulate my strategies.

Cloudscout – April 05, 2011 05:23PM Reply Quote
Det finnes ikke dårlig vær, bare dårlige klær!
I think you mean Fibber McGee & Molly's closet...

John Willoughby – April 05, 2011 05:59PM Reply Quote
Save us, Lord, from the furries of the Norsemen!
Yeah, thanks, I was going to look it up before clicking "post," but work required my attention and I let it go un-fact checked. Sorry. I don't think Amos and Andy even had a closet...

Cloudscout – April 05, 2011 07:41PM Reply Quote
Det finnes ikke dårlig vær, bare dårlige klær!
I'm an OTR nerd.

YDD – April 06, 2011 03:35AM Reply Quote

If possible, emphasize that you had no part in creating the situation
That bit's easy enough - CVS reveals that it was my boss who wrote the code in question a number of years before I arrived.

John Willoughby – April 06, 2011 07:57AM Reply Quote
Save us, Lord, from the furries of the Norsemen!
"Ha! It's not my fault, it's yours! And here's proof! What do you mean, I'm fired?"

ddt – April 06, 2011 08:15AM Reply Quote
Haha -- that was kind of my thought, too. You may KNOW it's his fault, HE may know it, but... .


YDD – April 06, 2011 09:01AM Reply Quote
To explain.... this is image registration software: take an image, and find the best transformation to match it to a generic image. The old version of this runs in about 60 minutes on a single core, and I was brought in to add GPU acceleration. Doing it the easy way, I got the time down to about 20 minutes, but there's a better way, which can do the whole thing in about 4 minutes. Unfortunately, the faster way gives a noticeably different transform (on about 75% of the images I've got). I've been trying to pin down the cause of the differences, and it became obvious quite quickly that the exact location of the transformed sample points was affecting things.[1,2] So I wound up trying to increase the number of sample points, so that a few 'bad' ones wouldn't affect the result so much. This is where the skeleton-in-the-closet turned out to be the admission attendant for the ossuary: increasing the number of sample points caused the original CPU code to give a significantly different answer (as in as different as the GPU results were). And ratcheting up the number of samples doesn't make things better (apart from demonstrating nicely how awesomely faster the GPU is at this sort of thing).

The upside is this: although they would like to ship in a week, the users are apparently trained to never mix analyses between releases - release 2.3 is expected to give significantly different answers to release 2.4. This is wrong for a host of other reasons (except, of course, if there was a definite bug fixed going 2.3 to 2.4 - in which case everything done with 2.3 needs to be re-run), but it does give some protection against changing things between releases.

[1] Boss: I thought I did bilinear interpolation on the pixel values!
Me: You did. After rounding the locations to integers, thereby rendering the interpolation pointless.

[2] And I think that my GPU method is noticeably more accurate

Edited 2 time(s). Last edit at 04/06/2011 09:02AM by YDD.

John Willoughby – April 06, 2011 09:43AM Reply Quote
Save us, Lord, from the furries of the Norsemen!
>After rounding the locations to integers


ARL (Moderator) – April 06, 2011 03:14PM Reply Quote
I whinge therefore I am!
Ah, the T.L.A.R.* system...

(That Looks About Right)

YDD – April 07, 2011 08:25AM Reply Quote
And another free titbit.... in this program, 'main' starts on line 192, and ends on line 1051. With another 1000 or so lines of other routines, it's actually one of the shorter ones in the package. The single file record is about 60k lines. And they don't believe in vertical whitespace.

John Willoughby – April 07, 2011 08:37AM Reply Quote
Save us, Lord, from the furries of the Norsemen!

John Willoughby – December 04, 2015 01:59PM Reply Quote
Save us, Lord, from the furries of the Norsemen!

ARL (Moderator) – December 04, 2015 02:53PM Reply Quote
I whinge therefore I am!
Speaking of the creator of the quote for this thread, anyone her about Ballmer's little meltdown at the MS shareholder meeting?

(He probably had a point, but well, not a tactful one...)

Edited 1 time(s). Last edit at 12/04/2015 02:54PM by ARL.

Cloudscout – December 04, 2015 04:49PM Reply Quote
Det finnes ikke dårlig vær, bare dårlige klær!
It's pretty much what everybody was thinking.


Past tense.

Since Ballmer said it would fail, everybody's now convinced that all of Microsoft's current strategies are going to be wildly successful.

Sorry, only registered users may post in this forum.

Click here to login