I joined and exposed the problem at: 07 Apr 2010


The problem was fixed in: July 26 2010



That is a ye..... wait, something is off in my maths

seems to be not handling GUIRWrappedText correctly...
Here is the data I used:

Code:

@Qazz: Yup, 2.5 months. You're welcome.

Thanks for the report, Eeems. I've added it to the To-Do List and will deal with it shortly.

thanks that will be mighty helpful

Resolved! It ended up taking about 10 bytes to repair the issue, once I tracked down how I could do so. Thanks for giving me a way to trace and test this with your AxeAid program.

Yay!
np

sweet!
downloading and uploading right now

Most superbulous, let me know what you find out. Current items on my To Do list:

To-Do Items

* DCS Menu item: Help
* Ans stack: sum(14) and sum(15)

Bugs

* Issue with some text not appearing in XLib programs? Cf. contrast text for Reuben: Ev by Kevin O.
* BASIC goto offset problems with XLib/Celtic?

Small gfx bug when DCS is first started, if you close the info pop-up, and then hit [2nd] the hourglass stays on the screen until you open the context menu(and I'm guessing the reg. menu as well)
Screenshot:


Little more severe bug that has survived since DCS6 and has caused the lcd to go into test mode for me once on hardware, is one described by the image below. It's a little more random, and it occurs if you run the program, close out of dcs, or even occurs sometimes when dcs first starts.

Weird, duplicated that one, no idea from where it's coming.

Another weird one, also duplicated. O_O

Edit: Fixed the first one. It was because of the FldSave routine getting triggered to back up your file structure on the first run. I moved that next to the backup DCS7 appvar getting created, so it's solved. On to your second bug..

Edit #2: A brief perusal is making me think that only unarchived programs are causing this instability, for some reason, which is the opposite of what I expected. The investigation continues.

*bump* Both repairs complete! As I said, the first bug was from the FolderBackup feature getting triggered. The second bug turned out to be from the BASIC Auto Upgrade feature trying to fix programs shorter than 8 bytes or so. The Auto Upgrade feature changes DCS headers that look like this:

Code:
into ones that look like this (note the quote):
Code:
It does the same for DCS6-headered BASIC programs, ie, those with 16x16-pixel icons. If the line with the icon was omitted, though, DCS would disturb memory in another program, possibly even its size word, when inserting memory, hence the random crashes, which only happened when DCS (or the TI-OS) would try to do certain things with programs in memory and the VAT. It didn't happen for archived programs with the same contents because archived programs aren't autoupgraded. Thanks for the excellent and very challenging bug reports, Tribal!

That's great you got them both fixed! If I find any more bugs, I'll be sure to post some

Thanks. As annoying and tedious as these bugs are to trace down, these are the little tiny kind of things that make people go from thinking a shell is rock-solid to not having complete confidence in it, so it's important to me to eliminate them. Cheers.

Bug:
Put this in a program:

Code:
Then run it. This crashes my calc with the most recent beta (the one with AnsStack).

Feature request:
On the "Error" screen during BASIC execution, could you make it so that keys other than 1 and 2 do nothing? Since the menu only lists those keys, and since I'm (and presumably other) so used to doing down->enter to "goto", this would be nice.

Did you happen to try the same with the DCS parser hook disabled?

Can do.

*bump*

Finally fixed. It's not the most attractive fix ever, but it works.

You got it.

Give it a try:
http://www.cemetech.net/files/dcs6.8.3_beta.8xk

I've got another one for ya:

Code:

And input only ". It will crash.
Also, a couple of issues with what I assume is whatever makes DCS run stuff at the homescreen. It really shouldn't auto clearscreen, sometimes you're doing stuff where you want what's on the home screen to still be there when you run it (like with little test programs). Also, Ans should persist when running a program. i.e.:

Code:
You should both continue to see the invocation of the program, and the output should be "AG1".

That's because Prompt doesn't modify Ans. If you do
Code:
then nothing bad happens as far as I can tell.

It doesn't do that for me, unless you run from the first line on 2.53MP. What OS, 1.19?

Edit: Actually, just tested 1.19, and it does what you said. I'm almost positive there's a reason I did that; let me look... Nope, no reason, and now I fixed it. Woot.

I see the invocation of the program, thanks to the fix above, but Ans is 0, how odd. Then again, what am I talking about? It's fine.

Alright, I tried your newest version, and it fixes the clearing the home screen issue. However, the Ans issue, and the pushing a null string onto the stack bug are still there.

Wait, pushing a null string? You said it was quote-dot, ". or "." if you prefer. Did I misunderstand?