Used to compile? (*solved*)

I think this used to work. Now it just gets a blank screen?

Based on HRPrintFast - as attached in the forum.

zxb --version
zxb 1.2.8-s758

http://dl.dropbox.com/u/4903664/HRPrintFast.zip

(I can't seem to attach this file to the forum either: Could not upload attachment to ./files/292_dbf5619d9a32ca38360a971b2a2970cd.)

[boriel]

Britlion forum attachment were disabled because I changed the forum to my own server.
please, try now

Also 1.2.8 should work.
Note: 1.2.8-s758 is in fact 1.2.9. This is the new version, so let me check it.
I think I accidentally overwrote the 1.2.8 branch with the new release. Let me restore it...

I just downloaded the beta version zxb 1.2.9-s765 and this also compiles to white screen.

The code ought to cls and then print 1 at least, I think ?

(since there's testing code that prints numbers each time it calls a subroutine to see how far it gets. It doesn't print any of them)

Okay. That's VERY strange.

I changed the start of the code to:

cls
print 0
pause 1
pause 0


And it did indeed pause. When I pressed a key, there was gentle sliding over the screen as expected.

So I deleted that code and recompiled. Boom. There was Mr. Gentle again.

I'm baffled by that behaviour...

I'm glad I added the .tzx to that zip file up there (I assume it doesn't do anything on your emulator?)

[boriel]

I'm trying to reproduce the bug and get the *minimal* code to isolate it.
If you can produce a smaller code, it would be of great help!

BTW Happy New Year!

Update: It seems your program is crashing due to out of memory. :S Removing the ALIGN 256 seems to solve it. Could be a bug in ALIGN 256 (still debugging).

Update 2: *Confirmed*. It's a bug in byte alignment.

[boriel]

I think I've fixed it. This could be a nasty and *dangerous* bug (thanks).
Please, download 1.2.9-s768 from the download page and tell mi if it works for you.

!



What bug did you find?

[boriel]

Under some circumstances (e.g. Label followed by an ALIGN statement), the program was misaligned, which means label addresses could be miscalculated, so instead of jumping or pokeing at an address, you could be jumping or pokeing at address - <offset error>
I've already uploaded a fix for 1.2.9 (1.2.8 fix still pending).

The nice thing is that the fix is just a line of code.