[gnu.gcc.bug] comparing stage2 with stage3 on sun386i

Leisner.Henr@XEROX.COM (Marty) (04/05/89)

I've been using gcc for a while but never followed the stage2/stage3
directions.

I finally tried it on  sun386i (gcc v1.33).  

I followed the directions in the manual for making stage1, 2 and 3 and ran 
cmp $file stage/$file

Byte 5 is different (which is the timestamp -- it makes sense -- I think).

I would assume non-coff systems don't contain the timestamp information in
the object files.

Anyway, the comment in the manual (section 2.10):
"Any difference, no matter how innocuous, indicates that the stage 2
compiler has compiled GNU C incorrectly..."
isn't really correct, at least for coff.

Guess a caveat should be added to the manual.

marty
ARPA:	leisner.henr@xerox.com
GV:  leisner.henr
NS:  martin leisner:wbst139:xerox
UUCP:	hplabs!arisia!leisner
 

blm@cxsea.UUCP (Brian Matthews) (04/07/89)

Marty (Leisner.Henr@XEROX.COM) writes:
|I've been using gcc for a while but never followed the stage2/stage3
|directions.
|I finally tried it on  sun386i (gcc v1.33).  
|
|I followed the directions in the manual for making stage1, 2 and 3 and ran 
|cmp $file stage/$file
|
|Byte 5 is different (which is the timestamp -- it makes sense -- I think).
|
|I would assume non-coff systems don't contain the timestamp information in
|the object files.
|
|Anyway, the comment in the manual (section 2.10):
|"Any difference, no matter how innocuous, indicates that the stage 2
|compiler has compiled GNU C incorrectly..."
|isn't really correct, at least for coff.

It's not only not really correct, it can be misleading.  As its default
action, cmp stops after it has found one difference.  You have to give it
a -l to show you all of the differences.  So when you did your cmp $file
stage/$file, and it told you byte 5 was different, it didn't tell you
whether or not any of the rest of the two files differ.

With coff object files, you should always do a cmp -l, and ignore
differences reported for bytes 5, 6, 7, and 8 (cmp reports differences by
counting the first byte of the file as 1.  Sigh) which is the timestamp.

-- 
Brian L. Matthews  blm@cxsea.UUCP   ...{mnetor,uw-beaver!ssc-vax}!cxsea!blm
+1 206 251 6098    Computer X Inc. - a division of Motorola New Enterprises