[comp.unix.ultrix] GCC 1.37.1 on a DECsystem 3100

alg@cornell.UUCP (Anne Louise Gockel) (06/05/90)

I followed the proceedures for compiling gcc 1.37.1 on a DECsystem 3100
running UWS 2.2 / Ultrix 3.1.  

The procedures suggest that you build gcc once with the Ultrix cc compiler, a
second time with the gcc compiler and a third time with the new gcc in order
to compare the binaries.  When I compared the binaries in the manner specified
in Step 10) of the INSTALL document, I found that many of the files had
differences beyond just the date stamp.  The differences were extremely
regular, almost as if there was a datestamp in the middle of the file.

Has anyone run into this problem?  Have I made a mistake in the installation
procedure?  Is there a good explanation for these differences?

					Thank you,
					Anne Louise Gockel
					Cornell Computer Science
Internet: alg@cs.cornell.edu
----------
Partial list of differences from the "cmp -l" command, comparing the stage2 and
the stage3 .o files: 
----------
c-convert.o
  4838  70  67
  4839  61  64
  4840  64  65
  4841  60  67
c-decl.o
 52282  70  67
 52283  61  64
 52284  62  64
 52285  70  65
c-parse.tab.o
 84669  70  67
 84670  61  64
 84671  62  63
 84672  62  71
c-typeck.o
 57886  70  67
 57887  61  64
 57888  63  65
 57889  64  61
caller-save.o
  4177  70  67
  4178  63  67
  4179  70  60
  4180  65  62
cccp.o
 90622  70  67
 90623  64  70
 90624  70  60
 90625  71  67
cexp.o
 12782  70  67
 12783  64  70
 12784  71  61
 12785  65  63
combine.o
 48006  70  67
 48007  63  66
 48008  64  66
 48009  71  63
cse.o
 52077  70  67
 52078  63  66
 52079  62  63
 52080  62  71
dbxout.o
emit-rtl.o
 16013  70  67
 16014  62  66
 16015  70  60
 16016  64  61

bogaart@serc.serc.nl (Eugene &) (06/05/90)

In article <41714@cornell.UUCP> alg@cornell.UUCP (Anne Louise Gockel) writes:


   The procedures suggest that you build gcc once with the Ultrix cc compiler, a
   second time with the gcc compiler and a third time with the new gcc in order
   to compare the binaries.  When I compared the binaries in the manner specified
   in Step 10) of the INSTALL document, I found that many of the files had
   differences beyond just the date stamp.  The differences were extremely
   regular, almost as if there was a datestamp in the middle of the file.

   Has anyone run into this problem?  Have I made a mistake in the installation
   procedure?  Is there a good explanation for these differences?

					   Thank you,
					   Anne Louise Gockel
					   Cornell Computer Science
   Internet: alg@cs.cornell.edu
   ----------
   Partial list of differences from the "cmp -l" command, comparing the stage2 and
   the stage3 .o files: 
   ----------
   c-convert.o
     4838  70  67
     4839  61  64
     4840  64  65
     4841  60  67
   c-decl.o
    52282  70  67
    52283  61  64

  etc...

I had the same problem with gcc 1.36, but I looked a bit further and
found that the so called differences are:

	diff ../testdiff/c-convert.o.hexl ./c-convert.o.hexl
	1c1
	< 00000000: 6201 0300 0eec 5826 3412 0000 6000 0000  b.....X&4...`...
	---
	> 00000000: 6201 0300 9df8 5726 3412 0000 6000 0000  b.....W&4...`...
	405c405
	< 00001940: 722f 746d 702f 6363 6130 3635 3130 2e73  r/tmp/cca06510.s
	---
	> 00001940: 722f 746d 702f 6363 6130 3334 3339 2e73  r/tmp/cca03439.s


According to me this is harmless, because they don't differ a lot !

The above output of a special diff shows that the first lines differ
and that half way the file is another difference. I have tested it,
and it appears to be in every *.o file. It is, I think, the name of
the intermediate assembler file where the *.o file is generated from
(but this is a guess !). Further I did not experience many problems in
using gcc (1.36) on the DEC Stations, but I did not use it extensively.

Eugene


--

Name:     Eugene Bogaart | Software Engineering Research Centre 
Email:   bogaart@serc.nl | Lange Viestraat 365  3511 BK Utrecht
Phone:	 +31 30 32 26 40 |           P.box 424  3500 AK Utrecht
Fax:     +31 30 34 12 49 |                      The Netherlands

teege@lan.informatik.tu-muenchen.dbp.de (Gunnar Teege) (06/05/90)

In article <41714@cornell.UUCP> alg@cs.cornell.edu (Anne Louise Gockel) writes:
>
>I followed the proceedures for compiling gcc 1.37.1 on a DECsystem 3100
>running UWS 2.2 / Ultrix 3.1.  
>... in Step 10) of the INSTALL document, I found that many of the files had
>differences beyond just the date stamp.  The differences were extremely

I had the same problem.
<bogaart@serc.nl> gave me the right tip: the files all differ in the
COFF-timestamp (as written in the INSTALL file) and somewhere half way
down, where they contain the name of the intermediate file used by the
compiler when they were generated (funny - of course these names are
different). I checked my files: All other parts are identical.

Does anyone know why these file names are included into the o-files
and which part is doing that?

Perhaps this could be eliminated. At least there should be inserted
a hint in the INSTALL document like that for the COFF timestamp.

Gunnar Teege   Inst. fuer Informatik, Technische Universitaet Muenchen
               PO-Box 202420, 8000 Muenchen 2,  West-Germany
teege@lan.informatik.tu-muenchen.dbp.de        +49 89 2105 8179
teege%lan.informatik.tu-muenchen.dbp.de@{unido.uucp,relay.cs.net,unido.bitnet}

meissner@osf.org (Michael Meissner) (06/06/90)

In article <2635@tuminfo1.lan.informatik.tu-muenchen.dbp.de>
teege@lan.informatik.tu-muenchen.dbp.de (Gunnar Teege) writes:

| In article <41714@cornell.UUCP> alg@cs.cornell.edu (Anne Louise Gockel) writes:
| >
| >I followed the proceedures for compiling gcc 1.37.1 on a DECsystem 3100
| >running UWS 2.2 / Ultrix 3.1.  
| >... in Step 10) of the INSTALL document, I found that many of the files had
| >differences beyond just the date stamp.  The differences were extremely
| 
| I had the same problem.
| <bogaart@serc.nl> gave me the right tip: the files all differ in the
| COFF-timestamp (as written in the INSTALL file) and somewhere half way
| down, where they contain the name of the intermediate file used by the
| compiler when they were generated (funny - of course these names are
| different). I checked my files: All other parts are identical.
| 
| Does anyone know why these file names are included into the o-files
| and which part is doing that?
| 
| Perhaps this could be eliminated. At least there should be inserted
| a hint in the INSTALL document like that for the COFF timestamp.

I replied to the original article in the proper newsgroup of
gnu.gcc.bug rather than the inappropriate group of gnu.gcc.  The short
answer is yes, the assembler is putting the intermediate filename in
the object file's file table, and the patches I just sent to the FSF
fix this through a kludge.

--
Michael Meissner	email: meissner@osf.org		phone: 617-621-8861
Open Software Foundation, 11 Cambridge Center, Cambridge, MA

Catproof is an oxymoron, Childproof is nearly so

frank@croton.enet.dec.com (Frank Wortner) (06/11/90)

As I remember it, there is a script called "ecoff-cmp" in the top level
directory of the GCC
distribution.  It strips out the headers and compares only the
instructions and data in the
executable file.  This has worked just fine the times I've tried it.

Regards,

					Frank