Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cornell!calvin!johns
From: johns@calvin.EE.CORNELL.EDU (John Sahr)
Newsgroups: gnu.gcc.bug
Subject: gcc on the DECstation 3100
Keywords: gcc-1.35 Ultrix 3.0 Decstation 3100
Message-ID: <1184@calvin.EE.CORNELL.EDU>
Date: 20 Jul 89 00:21:00 GMT
Organization: Cornell Space Plasma Physics, Ithaca, NY
Lines: 79

I have been trying to install gcc on our DECstations

DECstation 3100
Ultrix 3.0
gcc-1.35 from prep.ai.mit.edu

To configure, I ran "config.gcc mips"
then "make"

Three notable errors occur during the make.

1)  there was a slight error in tm.h, line 1274:

....
#define PRINT_OPERAND(FILE, X, CODE)  \
{ if ((CODE) == 'u')\
    if (TARGET_NOFIXED_OVFL)fprintf(FILE,"u");      \
  else if (GET_CODE (X) == REG)						\
    { extern char *reg_numchar[];\
      fprintf (FILE, "%s", TARGET_NAME_REGS ?reg_names[REGNO (X)]\
	       :reg_numchar[REGNO (X) ]);	\
    }\
  else \******** 1274 *********  the continuation '\' was missing
    { \

I fixed this, then typed "make", which proceeded unhindered until

2)  Many occurances of 

(from the compilation messages, first occurance...)

cc -c -g3   -I./config toplev.c
cpp: warning ./rtl.h:194: REG_P redefined

3)  Another error occured during the compilation of insn-emit.c

cc -g3  -I./config -c insn-emit.c
cpp: warning ./rtl.h:194: REG_P redefined
ccom: Warning: insn-emit.c, line 743: statement not reached
        operand0 = recog_operand[0];
      -----------------------------^

4)   Errors occurred in caller-save.c

cc -c -g3   -I./config caller-save.c
cpp: warning ./rtl.h:194: REG_P redefined
ccom: Error: caller-save.c, line 340: operands of : have incompatible types
            &&     (((regno) <32) ? (( DImode) != SFmode) && (( DImode) != DFmode)                : (   (( DImode) == SFmode) || (( DImode) == DFmode))                     && (((regno) % 2 ) == 0)))
      ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------^

similar type mismatches occurred at lines 343 346 349 and 352

5)  The make concluded with "`all' not remade because of errors"

Running "gcc -v" showed that gcc could at least report its version correctly.
However, upon attempting to compile the "hello world" program as

gcc hello.c -o hello

I got

gcc: installation problem, cannot exec cc1: No such file or directory

So, by the powers of deduction, I deduce that cc1 was not correctly made,
perhaps others as well.

Note that the INSTALLation instructions for config.gcc include the phrase 
"(But not the one from DEC)" for the "mips" argument, which I find curious,
with no obvious alternative mentioned. 

It is possible, of course, that I have done something desperately wrong.

Any clues, rumors or reports of success, patches, etc. will be greatly
appreciated.

john
-- 
John Sahr,     Dept. of Electrical Eng., Cornell University, Ithaca, NY 14853
johns@{alfven,calvin}.ee.cornell.edu,  {rochester,cmcl2}!cornell!calvin!johns