Hi All, there is a bug with `#include' in the new parser (only for orc
file or CsInstruments section in the csd file). The last line of a
file included through `#include' is analyzed two times by the lexer.
For example:

;; foo.txt
prints "=== FIRST  LINE ===\n"
prints "=== SECOND LINE ===\n"
prints "=== LAST   LINE ===\n"


<CsoundSynthesizer>
<CsInstruments>
#include "foo.txt"
instr 1
endin
</CsInstruments>
<CsScore>
i1 0 0
</CsScore>
</CsoundSynthesizer>


Tested with Csound 5.14, git commit 78d3db0ff720c7feec3914a20850aeeb75c084ef

uname -a
Linux pantera 2.6.33.7.2-rt30 #2 SMP PREEMPT RT Mon Oct 24 12:33:25 CEST 2011
x86_64 AMD Phenom(tm) II X6 1090T Processor AuthenticAMD GNU/Linux

csound -n include_test.csd
...
********************
* USING NEW PARSER *
********************
Testing...
...
<ast>
<tree type="292" name="T_OPCODE0" opname0="prints" >
 <treeright type="321" name="T_STRCONST" str=""=== FIRST  LINE ===\n"" >
 </treeright>
</tree>
<treenext type="292" name="T_OPCODE0" opname0="prints" >
 <treeright type="321" name="T_STRCONST" str=""=== SECOND LINE ===\n"" >
 </treeright>
</treenext>
<treenext type="292" name="T_OPCODE0" opname0="prints" >
 <treeright type="321" name="T_STRCONST" str=""=== LAST   LINE ===\n"" >
 </treeright>
</treenext>
<treenext type="292" name="T_OPCODE0" opname0="prints" >
 <treeright type="321" name="T_STRCONST" str=""=== LAST   LINE ===\n"" >
 </treeright>
</treenext>
<treenext type="301" name="T_INSTR" >
 <treeleft type="340" name="T_INTGR" value="1" >
 </treeleft>
</treenext>
</ast>
...
=== FIRST  LINE ===
=== SECOND LINE ===
=== LAST   LINE ===
=== LAST   LINE ===
orch now loaded
...


The cause is `PARM->buffer' (or yyextra->buffer) in csound_orc.l. The
scanner works well in the buffer state "yypushed" in the `do_include'
function, but after the `yypop_buffer_state', in the <<EOF>> rule,
PARM->buffer is still 'prints "=== LAST LINE ===\n"' and PARM->lBuffer
is 31 (the lenght of PARM->buffer), so this buffered line will be
analyzed again (in `get_next_char' used by `YY_INPUT'). A temporary
solution is


diff -ur csound5~/Engine/csound_orc.l csound5/Engine/csound_orc.l
--- csound5~/Engine/csound_orc.l	2011-11-08 13:51:46.000000000 +0100
+++ csound5/Engine/csound_orc.l	2011-11-12 17:36:27.000000000 +0100
@@ -432,6 +432,8 @@
                   yypop_buffer_state(yyscanner);
                   if ( !YY_CURRENT_BUFFER ) yyterminate();
                   printf("End of input; popping to %p\n", YY_CURRENT_BUFFER);
+                  if (PARM->isInclude)
+                    PARM->lBuffer = PARM->isInclude = 0;
                   n = PARM->alt_stack[--PARM->macro_stack_ptr].n;
                   printf("n=%d\n", n);
                   if (n!=0) {
@@ -530,6 +532,7 @@
     }
     PARM->alt_stack[PARM->macro_stack_ptr].n = 0;
     PARM->alt_stack[PARM->macro_stack_ptr++].s = NULL;
+    PARM->isInclude = 1;
     yypush_buffer_state(yy_create_buffer( yyin, YY_BUF_SIZE, yyscanner), yyscanner);
 }
 
diff -ur csound5~/Engine/parse_param.h csound5/Engine/parse_param.h
--- csound5~/Engine/parse_param.h	2011-11-08 13:51:46.000000000 +0100
+++ csound5/Engine/parse_param.h	2011-11-12 17:36:45.000000000 +0100
@@ -30,6 +30,7 @@
     unsigned int macro_stack_ptr;
     char            *xstrbuff;
     int             xstrptr,xstrmax;
+    unsigned char   isInclude;
 } PARSE_PARM;
 
 #define lMaxBuffer (1000)


The `if' condition is necessary, otherwise the $NAME macro doesn't work.

This is my first message, my English is *underground* but I know how to say

  Thank you very much for Csound

tito

------------------------------------------------------------------------------
RSA(R) Conference 2012
Save $700 by Nov 18
Register now
http://p.sf.net/sfu/rsa-sfdev2dev1
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net