#4045 Three mcs51 large & huge const propagation failures

[Maarten Brock]

In the regression tests three tests fail with a crash in SDCCgenconstprop.cc line 1402 for mcs51 large & huge model.

/home/sdcc-builder/build/sdcc-build/orig/sdcc/support/regression/cases/../tests/gcc-torture-execute-990804-1.c:39: error 9: FATAL Compiler Internal Error in file '/home/sdcc-builder/build/sdcc-build/orig/sdcc/src/SDCCgenconstprop.cc' line number '1402' : code generator internal error 
Contact Author with source code
gmake[5]: *** [Makefile:437: gen/mcs51-large/tst_gcc-torture-execute-990804-1.rel] Fehler 1
results/mcs51-large/tst_gcc-torture-execute-990804-1.out:2:--- FAIL: cannot compile/link cases/tst_gcc-torture-execute-990804-1.c

/home/sdcc-builder/build/sdcc-build/orig/sdcc/support/regression/cases/../tests/gcc-torture-execute-divmod-1.c:14: error 9: FATAL Compiler Internal Error in file '/home/sdcc-builder/build/sdcc-build/orig/sdcc/src/SDCCgenconstprop.cc' line number '1402' : code generator internal error 
Contact Author with source code
gmake[5]: *** [Makefile:437: gen/mcs51-large/tst_gcc-torture-execute-divmod-1.rel] Fehler 1
results/mcs51-large/tst_gcc-torture-execute-divmod-1.out:2:--- FAIL: cannot compile/link cases/tst_gcc-torture-execute-divmod-1.c

cases/onebyte/onebyte_attrL_volatile_attrR_none.c:40: error 9: FATAL Compiler Internal Error in file '/home/sdcc-builder/build/sdcc-build/orig/sdcc/src/SDCCgenconstprop.cc' line number '1402' : code generator internal error 
Contact Author with source code
gmake[6]: *** [Makefile:437: gen/mcs51-large/onebyte/onebyte_attrL_volatile_attrR_none.rel] Fehler 1
results/mcs51-large/onebyte/onebyte_attrL_volatile_attrR_none.out:2:--- FAIL: cannot compile/link cases/onebyte/onebyte_attrL_volatile_attrR_none.c

[Philipp Klaus Krause]

I can reproduce the issue on my Debian GNU/Linux testing system using SDCC built from current svn trunk.

[Philipp Klaus Krause]

The quick and easy way would be to just re-disable the assertion for mcs51. The better way would be to find out why we get the corrupted iTemp usage information for mcs51 in the first place.

[Philipp Klaus Krause]

Small code sample to reproduce the issue:

int gfbyte ( void ) 
{
 return 0;
} 

void
f (void)
{
 int i,k ;

 i = gfbyte();

 i = i + 1 ;

 if ( i == 0 ) 
     k = -0 ;
 else
     k = i + 0 ;

 if (i != 1)
   return;
}

[Philipp Klaus Krause]

This case (and gcc-torture-execute-990804-1) is fixed in [r16724]. It was an issue in deleteIfx in SDCClabel.c corrupting OP_USES.

The other two tests still fail.

[Philipp Klaus Krause]

Last part of the bug fixed in [r16725]. CSE could introduce iCodes not in iCodehTab.