unexpected char leads to flooding about bitlash HOT 16 CLOSED

Thanks for the report. I believe the same fix that addresses issue #3 may address this one. I tested on Ubuntu, Screen version 4.00.03jw4. Code is in commit b612b7 for src/bitlash-cmdline.c. I'll leave this issue open in the hope you'll update it once you get a chance to test. Thanks,

-br

from bitlash.

the issue persists both when I type and when I ctrl-v.
Also, if for example (by mistake) you put a html link the same happens (pointing at the : )

1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345line overflow

e �
( �
( �
( �
( �
( �
( �
(

as

---^
unexpected number

#2

-------^
unexpected number

t �
( �
( �
( �
(

from bitlash.

Thanks for the update. Well that is odd. With the patch applied, there is no code anywhere in Bitlash to generate a Line overflow. Could you please post the contents of your copy of src/bitlash-cmdline.c?

from bitlash.

you still have overflow on:
bitlash-parser.c:58: "expected \0unexpected \0missing \0string\0 underflow\0 overflow\0^C\0^B\0^U\0exp\0op\0:xby+-*/\0eof\0var\0number\0)\0saved\0eeprom\0:="\0> \0line\0char\0stack\0startup\0id\0prompt\0\r\nFunctions:\0oops\0arg\0function\0"

#include "bitlash.h"

// Serial command line buffer
char *lbufptr;
char lbuf[LBUFLEN];

// Help text
//
#ifdef ARDUINO_BUILD
const prog_char helptext[] PROGMEM = { "http://bitlash.net\r\nSee LICENSE for license\r\nPins: d0-22,a0-22 Variables: a-z, 32 bit long integers\r\nOperators: + - * / ( ) < <= > >= == != << >> ! ^ & | ++ -- :=\r\nCommands: \0" };
#else
const prog_char helptext[] PROGMEM = { "http://bitlash.net\r\n\0" };
#endif

void showdict(const prog_char *addr) {
byte c;
for (;;) {
c = pgm_read_byte(addr++);
if (c == 255) return;
else if (c == 0) {
if (pgm_read_byte(addr) == 0) return;
else spb(' ');
}
else if (c != '`') spb(c);
}
}

void displayBanner(void);

void cmd_help(void) {
displayBanner();
showdict(helptext);
showdict(reservedwords);
msgp(M_functions);
#ifdef LONG_ALIASES
showdict(aliasdict);
speol();
#endif
showdict(functiondict);
speol();
show_user_functions();
speol();
void cmd_ls(void);
cmd_ls();
}

void prompt(void) {
char buf[IDLEN+1];
// Run the script named "prompt" if there is one else print "> "
strncpy_P(buf, getmsg(M_promptid), IDLEN); // get the name "prompt" in our cmd buf
if (findscript(buf)) doCommand(buf);
else msgp(M_prompt); // else print default prompt
}

void initlbuf(void) {
lbufptr = lbuf;

#if defined(SERIAL_OVERRIDE) && 0
// don't do the prompt in serialIsOverridden mode
if (serialIsOverridden()) return;
#endif

prompt();

// flush any pending serial input
while (serialAvailable()) serialRead();

}

// Add a character to the input line buffer; overflow if needed
byte putlbuf(char c) {
if (lbufptr < lbuf + LBUFLEN - 2) {
*lbufptr++ = c;
spb(c);
return 1;
}
else {
spb(7); // beep
return 0;
}
}

void pointToError(void) {
if (fetchtype == SCRIPT_RAM) {
int i = (char *) fetchptr - lbuf;
if ((i < 0) || (i >= LBUFLEN)) return;
speol();
while (i-- >= 0) spb('-');
spb('^'); speol();
}
}

// run the bitlash command line editor and hand off commands to the interpreter
//
// call frequently, e.g. from loop(), or risk losing serial input
// note that doCommand blocks until it returns, so looping in a while() will delay other tasks
//

///////////////////////
// handle a character from the input stream
// may execute the command, etc.
//
void doCharacter(char c) {

if ((c == '\r') || (c == '\n') || (c == '`')) {
    speol();
    *lbufptr = 0;
    doCommand(lbuf);
    initlbuf();
}
else if (c == 3) {      // ^C break/stop
    msgpl(M_ctrlc);
    initTaskList();
    initlbuf();
}
else if (c == 2) {          // ^B suspend Background macros
    suspendBackground = !suspendBackground;
}
else if ((c == 8) || (c == 0x7f)) {
    if (lbufptr == lbuf) spb(7);        // bell
    else {
        spb(8); spb(' '); spb(8);
        *(--lbufptr) = 0;
    }
} 

#ifdef PARSER_TRACE
else if (c == 20) { // ^T toggle trace
trace = !trace;
//spb(7);
}
#endif
else if (c == 21) { // ^U to get last line
msgpl(M_ctrlu);
prompt();
sp(lbuf);
lbufptr = lbuf + strlen(lbuf);
}
else putlbuf(c);
}

/////////////////////////////
//
// runBitlash
//
// This is the main entry point where the main loop gives Bitlash cycles
// Call this frequently from loop()
//
void runBitlash(void) {

// Pipe the serial input into the command handler
if (serialAvailable()) doCharacter(serialRead());

// Background macro handler: feed it one call each time through
runBackgroundTasks();

}

from bitlash.

Thanks for the quick reply. That is the correct code, so we still have a puzzle.

It's true that the strings "overflow" and "line" are still in place, but at least here I cannot find any remaining calls to "overflow(M_line)" after the changes I just made to bitlash-cmdline.c. You might grep the src/ directory to double-check me. So the patched version of Bitlash should not be able to emit that line overflow that you are observing. Other overflows, of course.

I hate to theorize like this but one has to ask: Is it possible that you somehow have a stale binary on the Arduino? Because I can't figure out how else Bitlash can say "line overflow". It doesn't do that any more, with the patch.

If not that, I'm short of theories. Any ideas at your end? I should add this all works happily here:

bitlash here! v2.0 RC5pre (c) 2011 Bill Roy -type HELP- 973 bytes free

print 0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000001
1

One random thought that is worth a quick check: does it help if you comment out the call to traceback() at line 47 in src/bitlash-error.c?

Any further thoughts/data/ideas welcome.

-br

from bitlash.

You are completely correct. After shutting down arduino and re-compiling now there's no error anymore.
Perfect!
Congratulations and Thanks

In the same line, is it possible to have a multi line command?

Also,
if I put an weird command such as a link:
#2 (comment)

I get the result:

#2 (comment)

-------^
unexpected number

tBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB

from bitlash.

Glad we could corner and eliminate that one. Whew.

The parser is actually capable of handling multi line input text - though the SD card example shows it sometimes takes great lengths. But you can't type such commands in using the built-in line editor. I am guessing you want a multi line input editor, which would indeed be nice. And I accept code contributions...

I'm still puzzled by the character spew you are seeing after the error message text is printed. It is happening right where the traceback is supposed to be printed, and I suspect the bug is there. If you haven't already done so, comment out the call to traceback() at line 47-ish in src/bitlash-error.c to prevent the traceback from being printed, and let me know if the behavior is any better. Thanks.

-br

from bitlash.

yes. no traceback.
any guidance on where could i implement the multiline?
although not 100% fluent in c++ (i work mostly with python and perl), I think I can help.

from bitlash.

If traceback is off then I'm very puzzled again about the runaway printing after the error. Could it be a free ram issue, I wonder? How much free ram do you have at startup? I will give this a further think.

Regarding multiline: The current single-line editor is in the file we're looking at now: bitlash-cmdline.c; that's where you might start to think about multiline, perhaps replacing doCharacter() with something smarter. The editor's job is to assemble a null-terminated buffer of text and pass it to doCommand(). It currently uses one fixed-length buffer, linebuf, to do that. You could make linebuf bigger, for starters, but do mind your free ram.

You won't find much C++ in Bitlash; it's mostly C for space reasons. I don't think it will be difficult if you already are comfortable in Python and Perl.

-br

from bitlash.

Just a thought: Do you have a "prompt" function? That's the next thing after the traceback.

from bitlash.

created by myself overriding the system's one?
no

from bitlash.

I've had no new inspirations overnight about a possible cause for the runaway output.

Maybe the "const prog_char" changes have an impact somehow? Anyway, could you post your bitlash directory to github so I can pull it down and try to reproduce the problem here?

from bitlash.

I've deleted it all and pull from your site. no mods

from bitlash.

I believe this is fixed in commit 5cfc04b. Please let me know.

from bitlash.

problem fixed.
now if I type a wrong command, only a warning appears. No need to reset anymore.

Thanks

from bitlash.

Glad to see it's fixed; closing this issue.

from bitlash.