"Tone" problem

[Anobium]

What are the changes you make ?
Remind which OS.

---

An insight.

To build all the distributions of Great Cow BASIC, and, now the updates are fully automated here.

• R2Build. Software that can execute 100s of steps in sequence, from simple ini file updates to complex database changes.
• OS Command files. Using R2Build I have 11 small batch files that do things at the OS level like rename all files to lower case (needed for the Linux build).

I will gladly share the tools I use, and, the processes to help.

[mkstevo]

The update automatically overwrites a lot of the .ini and .bat files that I then have to try and remember which files I need to change back again to return the NSDSP programmer and Geany to working order.

Right now at home I'm using Windows 10 but at work I'm still on my iMac. I don't have any inclination to start on updating the Mac version again. Especially as I no longer have a Mac at home.

[stan cartwright]

Yeah source forge eventually does download gcb but takes ages... never did before.

[mkstevo]

The download took a few seconds for me?

[stan cartwright]

Do you think it has taken so long to find errors in gcb because no one uses the commands.
I would think, as a beginner, that sound would be interesting...but then you got to have a "speaker"...so too much bother.
Can't do much with a ucontroller unless you connect it to something.
GCB does that easyish.
We aren't all pic users. I'm liking logic green technology....the name sounds cool and it's zippy with gcb.

[Anobium]

Do you think it has taken so long to find errors in gcb because no one uses the commands.

No. FOR-NEXT is use extensively. The issue was deep in the compiler and we did not pick up on testing earlier this year.

[William Roth]

@Stan

A bit of history. "For next" had issues for a very long time. But these would not show up when stepping with a value of 1, which is the most common usage. However, when stepping with other values (2 for example) the for next loop could become endless as overflows were not being handled correctly in certain cases. This was all fixed in one of the release candidates.

However there was a typo in a section of the new for next code. This typo was responsible for the issue that JackJames had with the Tone code. Under somewhat rare circumstances this typo caused a variable used with for next not to be cast correctly . This problem did not show up under testing before release of the code. It has been fixed now in the latest build.

The latest build is: 0.98.07 2021-07-24 : Build 1008

https://sourceforge.net/projects/gcbasic/files/Release%20Candidates/Patches

This should be considered a "Critical Update" and as a minimum going forward. Yes, updating may cause some time and inconvenience. However, without this update the generated ASM is likely to be incorrect and certain programs or libraries may not operate as expected.

---

If you care to see an example of the original for next issue try this code:

 #Define uselegacyfornext

  Dim CntA as byte

For CntA = 0 to 255 step 2
      Hserprint CntA
      HSerprintCRLF
 Next

The for next will loop endlessly.
Remove the #define and it will execute correctly

Edited link to point to the patch to update to Build 1008

[stan cartwright]

You explain the problem well @William Roth.
Never heard off #Define uselegacyfornext

It's the sort of thing the user would have seen.. ie an overflow.
The fact that gcb didn't do something else when compiling like a check is an obvious oversight but can see now how it got this far without being noticed.
My reference to tone was who uses it?
There's some dma stuff on the forum which looks interesting but probably hard to understand pic stuff.
I got to get into pics more but find it easier to use avr cos they come on boards with lots of pins and the usb which you only get with pic dev boards which cost more.
I hope that avr support.. like the lgt328 is kept up as it's great for a lot of stuff and is now sorted and consistent .
Don't get me wrong. I think the pic stuff is brill.. but sometimes over my head.
The idea of a pic with an op amp built in seems a good idea but in practice I bet it's not trivial to set up. So I'd go for a discrete op amp and a bit of extra circuitry.
Still think gcb is really good software.

[stan cartwright]

So the for next problem is the step... which is sorted.
so for count1=0 to 255 step 3 would work but
for count1=0 to 255 step 6 wouldn't.
edit is that what the problem was and did it apply to word value counters as well?
I wouldn't understand how it was resolved but did it test if the step value /255 or /65535 was an exact number ie no remainder? Just a thought.

[Anobium]

@Stan.

The problem was for all variable types. The problem is described with the solution here, see https://sourceforge.net/p/gcbasic/discussion/596084/thread/59d24215/#b174 (same post as Bill Roth shared)

The solution was not as simple as shown on the post above but is gave me insight to resolve.

The original code was simple and covered most cases. But, as the URL above discusses the loops are unacceptable. There is a trade off. The legacy code is smaller ( in this example ) using90+ words of program memory to do all the additional checking.

More example code. If you try this on an UNO you will see the issue.. just uncomment #DEFINE USELEGACYFORNEXT to see the continuous loop.

#chip 18F26K22, 4
#OPTION Explicit

'#DEFINE USELEGACYFORNEXT


  'USART settings
  #define USART_BAUD_RATE 9600
  #define USART_TX_BLOCKING
  #define USART_DELAY 10 us

  dim count1 as Byte

  HSerPrintCRLF 2
  HSerPrint "Count step 3"
  HSerPrintCRLF
  For count1 = 0 to 255 step 3
    HSerPrint str(count1)
    HSerPrint ","
  Next

  HSerPrintCRLF 2
  HSerPrint "Count step 6"
  HSerPrintCRLF
  For count1 = 0 to 255 step 6
    HSerPrint Str(count1)
    HSerPrint ","
  Next

  HSerPrintCRLF
  wait 10 ms
  Do
  Loop

[stan cartwright]

This for next error wouldn't have become apparent without sloppy. deliberate programming.
When you think about it it's obvious it could happen... just one of those things that never happens.
Well it never happened to me.. luck?

[William Roth]

Posted by Stan

This for next error wouldn't have become apparent without sloppy. deliberate programming.

This is simply not true. And, likely based upon how you personally presume a program should be constructed . A perfectly well constructed program could fail under the right conditions. I provided one example.

A For Next has a start value an end value and a step value. Many times we use constants for these since we know where to start, where to end and what the step value should be. But this is not always the case. If these are variables that could change then the compiler needs to be able to handle these in a proper manner without generating final code that is unpredictable.

I can think of several cases where the original for next could fail without "sloppy programming". Perhaps you should educate yourself a bit more before making such off-hand comments.

Perhaps reading this thread might help you.

https://sourceforge.net/p/gcbasic/discussion/596084/thread/59d24215/#b174

William

[mkstevo]

I seem to have scuppered a previously working program.

I have a For - Next loop which has a negative step value. The value is passed as a parameter into the subroutine. I'm getting an error which states:

Error: Missing closing double quote in Table source:221 '" 
Current/C1_16f1829_3.04.gcb (651): Error: For-Next 'step' must 
be Integer variable

Sub DoCountDownSub(Start As Word, Reduce As Byte)

  If Start > 9999 Then
     Let Start=9999
  End If


  For LoopCnt=Start to 0 Step - Reduce
      ShowValuesSub(LoopCnt)
      ...
      'Rest of loop here
  Next

Ignoring the Table error for a second, should I be creating the step variable as a separate byte variable within the subroutine and then setting that variable to the value of my parameter or should I be able to use the byte value parameter directly as I was doing previously?

[Anobium]

For LoopCnt=Start to 0 Step - Reduce

The variable Reduce needs to an integer as the error states.
'Reduce As Integer' is needed with the FOR-NEXT routines.

---

If someone can figure out how NOT to have an integer then I would be grateful for the help. But, as you are using a variable the type must be an integer.

[mkstevo]

Hmm. I think I've missed the point somewhere.

Are you saying that my 'Byte' parameter should be changed to an 'Integer' parameter?

I've never used an integer in GCB. Bytes, Words, Longs yes. Integers? No!

[mkstevo]

Re-writing the parameter so that it reads:

Sub DoCountDownSub(Start As Word, Reduce As Integer)

Isn't the answer.

Nor is:

Sub DoCountDownSub(Start As Word, Reduce As Integer)

Dim ReduceVar As Integer
Let ReduceVar = Reduce

  If Start > 9999 Then
     Let Start=9999
  End If


  For LoopCnt=Start to 0 Step - ReduceVar

[Anobium]

Please use this as a workaround

I negate the ReduceVar then use that negated value.

#chip 16f1829
#option Explicit

dim vari as Byte
dim stepvar as Byte

DoCountDownSub ( vari, stepvar )

Sub DoCountDownSub(Start As Word, Reduce As Integer)

Dim LoopCnt as Word

Dim ReduceVar As Integer
ReduceVar = -Reduce

  If Start > 9999 Then
     Let Start=9999
  End If


  For LoopCnt = Start to 0 Step ReduceVar

  Next

End Sub

There is something not correct in the compiler. It is probably just an errant message but I will have to investigate to resolve. I did ask people to test the FOR_NEXT changes as these changes were extensive - never enough testing :-(

[mkstevo]

That compiles, thanks for your thoughts.

{Further comments deleted as, again, they might have come across as snarky. Apologies once more.}

[mkstevo]

Answering my own confusion here.

{I had some other comments regarding use of constants here. I later removed those comments as it has been suggested that although the program compiled it might not have worked as intended.}

I tried leaving the program as it was, adding the legacy directive:

Sub DoCountDownSub(Start As Word, Reduce As Byte)



  If Start > 9999 Then
     Let Start=9999
  End If

#Define UseLegacyForNext

  For LoopCnt=Start to 0 Step - Reduce

This resulted in a successful compile.

[Anobium]

Answering my own confusion here. The value for the 'Step' must be a fixed constant now. Not a parameter (of any type), not a variable (of any type), a fixed value constant.

No! Like I said in this post https://sourceforge.net/p/gcbasic/discussion/596084/thread/a56563e36f/#c0fb/747c/4310/d6f8

Your analysis is incorrect.

Please look at https://sourceforge.net/p/gcbasic/discussion/596084/thread/a56563e36f/#c0fb/747c/4310/d6f8

[mkstevo]

Deleted.

Although I was trying to help with my replies, I don't think they helped at all.

My efforts in trying to locate the cause of the problem may have been interpreted as criticism. That was not my intention and I apologise.

[Anobium]

You analysis is correct. Using constants is not the resolution. It is a theory but analysis show it is another cause.

---

The issue is the NEGATing of the value '-Reduce'. This is causing the variable Reduce to be treater as a BYTE. This has nothing to do with the new FOR-NEXT code as the negate happens before the FOR-NEXT section of the compile is called.

  For LoopCnt = Start to 0 Step -Reduce
  Next

Proof (I have added debug but this shows this to be the case. This casts '-Reduce' to an integer.

  For LoopCnt = Start to 0 Step [INTEGER]-Reduce
  Next

As the negate causes the variable to be a byte the ASM generated contains no step variable decrement step code (just increment step code).

This needs fixing.

[mkstevo]

The table error is easily fixed, though a little fussy.
The error refers to:

Error: Missing closing double quote in Table source:221 '" 

I have a lookup table for the alphabet which maps ascii codes to some values which cross reference a first (alphabetic) table into the numeric values needed to show that character on four seven segment LED displays driven by four 74HC595.
221 is a value which lights the top two vertical segments, to give the effect of a double quote. My table is (or was) correct, the quote symbol is 'Rem'ed out by the preceeding single quote.

I can replace the double quote with the phrase 'Double Quotation mark' so it can soon be repaired, though it is a little odd that a remark should be attempted to be compiled?

Table Alphabet As Byte
255 '  Space
255 '! Space
221 '" 
255 '# Space
255 '$
255 '% Space
255 '& Space
223 ' 
198 '( Square bracket
240 ') Square bracket
255 '* Space
255 '+ Space
127 ', Single decimal point
191 '- Dash in the middle of the display
127 '. Single decimal point
237 '/ Two offset vertical bars 'i
192 '0
249 '1
164 '2
176 '3
153 '4
146 '5
130 '6
248 '7
128 '8
144 '9
255 ': Space
255 '; Space
167 '< 'c'
183 '= Two dashes, one at the bottom, one in the middle
179 '> 'c' inverted
172 '?
255 '@ Space
136 'A
128 'B
198 'C
192 'D
134 'E
142 'F
194 'G
137 'H
207 'I
241 'J
137 'K
199 'L
200 'M
200 'N
192 'O
140 'P
152 'Q 'q'
206 'R
146 'S
135 'T
193 'U
193 'V 'U'
193 'W 'U'
137 'X 'H'
145 'Y
164 'Z looks like a 'two'
198 '( Square bracket
219 '\ Two offset vertical bars i'
240 ') Square bracket
220 '^ 'n' in top portion of display
247 '_ Dash at the bottom of the display
253   '
163 'a 'o'
131 'b
167 'c
161 'd
134 'E
142 'f
144 'g
139 'h
239 'i
241 'j
139 'k
207 'l
171 'm
171 'n
163 'o
140 'p
152 'q
175 'r
146 'S
135 't
227 'u
227 'v 'u'
227 'w 'u'
137 'x 'H'
145 'y
164 'z looks like a 'two'
198 '{ Square bracket
207 'l Two vertical bars to the left of the display 
240 '} Square bracket
254 '~ Dash at the top of the display
End Table

[Anobium]

We have the same issue in the major tables in the compiler. This is not resolvable as the double is the string delimiter.

Your method to resolve is the way to resolve.

[mkstevo]

Thanks. Though it had been working without issue previously.