Spasm Linux Errors

[NanoWar]

With GNU-Debugger I tracked runtime errors:

Code: Select all

Program received signal SIGSEGV, Segmentation fault.
0x0000000000408917 in handle_bitmap(_IO_FILE*, tagRECT const*, tagBITMAPFILEHEADER const*, tagBITMAPINFOHEADER const*, tagRGBQUAD*) ()

(Compiling Wizards)

[Buckeye Dude]

Wanna send me your code for assembling wizards? We dont have a lot of tests dealing with bitmaps, we probably should. That code needs cleaning up anyway hopefully this will be an excuse to do so.

[NanoWar]

It segfaults on any bitmap inclusion. Trivial example is in the attachment.

[Buckeye Dude]

Hmm confirmed, athough I'm not sure whats causing this. Theres no issue on windows, and Spencer says the mac version doesn't have a problem either. Guess its time for some command line gdb :/

[NanoWar]

How do you professionally debug a C program in Linux? I dont think GDB command line is any fun...

You could use an IDE I would think. Eclipse and KDevelop are two good ones.

[Buckeye Dude]

Most people on linux have there own system of doing things, IDEs are probably the easiest way, but most people are already using vi or emacs, which both have plugins for using debugging tools I believe. And yeah at first I was skeptical of command line debugging but actually its very doable, although a little slower than stepping through with an IDE.

[Buckeye Dude]

Got it. LONG was incorrectly defined as unsigned and it was stuck in an infinite loop. Checkout from svn and rebuild, or download the latest off codeplex

Works for me

[NanoWar]

Hey, got another one with the seek directive!

Code: Select all

#define TI83P
#include "ti83plus.inc"

      .org $9D95
      .db $BB,$6D
start:
      ld hl, 1234
      bcall(_DispHL)
      bcall(_newLine)
      ret
#define save eval($)
      .seek start
      ret ; crashHere
      .seek save

Code: Select all

Program received signal SIGSEGV, Segmentation fault.
0x0000000000402d54 in write_out (val=123) at pass_one.cpp:66
66          *(out_ptr++) = val;

We can advance the errors only by one at a time it seems. I am trying to compile Wizards which uses tons of file input/output in macros which is where Unix likes to segfault often I guess.

[Buckeye Dude]

Another sign issue with 64 bit. Fixed, you can checkout and rebuild

[NanoWar]

I have more

Code: Select all

; define a macro
#macro func( func_parameter )
      #ifdef func_parameter
            .echo "defined :)"
      #else
            .echo "undefined :("
      #endif
#endmacro

; wrap it
#define .wrap_func func(

; call them
      func( 1 )
      func()
      .wrap_func 1
      .wrap_func ; # ERROR!
 

Another one:

Code: Select all

      .org 0
start
      .db '0'
save
      .seek start
      .db '1'
      .echo "1 start: ", start
      .echo "1 save: ", save
      .echo "1 exit: ", exit ; # ERROR!
      .seek save
exit
      .echo "2 start: ", start
      .echo "2 save: ", save
      .echo "2 exit: ", exit
 

And:

Code: Select all

#macro func( param )
      #if param = -1
            .echo "Hmm?"
      #endif
      .db param
      .echo param ; # UNEQUAL!
#endmacro

      .db -1
      func(-1)
 

Also the APP signing went wrong again. DIFF on my .bins is successful, on .8xk it's not (comparing to an old build of spasm).

[Buckeye Dude]

Fixed the app signing as you know. First and third are also fixed, second is expected, you can't forward ref a label from an echo, its evaluated whereever its encountered during the first pass. Committed and released on codeplex

At least you are probably going to catch most of the errors

[Buckeye Dude]

I wish, I'm sure there are tons more. And our unit test coverage is absolute crap, to release it just has to assemble Zelda. Too bad unmanaged code in VS doesn't have the nice code coverage features of managed, if so I'd have that it 100% covered.

[NanoWar]

You could use my example snippets to ensure that they work in coming revisions. You should verify all output, maybe compare binaries also with intel hex...

[Buckeye Dude]

Oh yeah, we do regression testing for pretty much all broken code, however writing more tests is a pain, especially since VS doesn't support unit tests for C++. And without an analysis of whats covered its hard to motivate myself to figure out what needs covered and write tests to cover it

[NanoWar]

I just downloaded the latest build, and the Wizards output files are again different!

Code: Select all

#macro function(value)
      #ifndef value
            #define value 0
      #endif
      .echo value
#endmacro

#define .function function(

      .echo "Expected: 1"
      .function 1
     
      .echo "Expected: 0"
      .function
 

...does not work anymore (again!).

Code: Select all

crash6.z80:14: Error during invocation of macro '.FUNCTION'
crash6.z80:8: error SE104: Expecting a value, expression ended early

[Buckeye Dude]

You're positive you're using the version I fixed? I can't test on linux at the moment but I ran in Windows and it worked fine

[NanoWar]

Yes, sure, I've got my dream build already