From:	Andrew Crabtree <andrewc AT typhoon DOT rose DOT hp DOT com>
Message-Id:	<199705202219.AA239306756@typhoon.rose.hp.com>
Subject:	PGCC Update
To:	djgpp AT delorie DOT com
Date:	Tue, 20 May 1997 15:19:15 PDT

Hi Guys - 

2 things about PGCC.

1)  I originally only told people to add -Wno-sign-compare after -Wall to 
get it to behave like old gcc-2.7.2.  I have now learned that 
-Wimplicit is also enabled by default in the new snapshots. 

So - 

For gcc-2.7.2 

-Wall

is (roughly) equivalent to 

pgcc 

-Wall -wno-implicit -Wno-sign-compare

This makes recompiling libc much easier.


2)  The stupid alignment problem.  

I still don't know exactly what this is.
Through experimentation I have found out that the DJGPP as will accept align
values from 1-15.  PGCC outputs align values of 16 for -m486 and -mpentiumpro
which will stop your compilation (align value too large, 15 assumed).
I checked the as that I built as a 
cross-assembler from unix and it behaves differently.  It only allows align
values of 1,2,4,8, and 16, and will complain about anything else ( .align 3, which djs
accepts, errors with 'not a power of 2').  I rebuilt
as for DJGPP from the bin-utils2.7 sources and (my copy anyway) it now works on the
 1,2,4,8,16 scale.  Doing
an as -v on both of them yields the exact same results, so I assume they were
built from the same sources. I am at a loss to explain the difference.  
I'll put together a new zip file containing just the replacement as.exe, and
it should appear on the pcg pages just as soon as I tell them about it.  It will not
be tested very well, so hackers only.  If anyone knows a way that I can 
inspect the coff output of the assemblers to see exactly what is going on 
I would appreciate the knowledge.  Also, if anyone knows more about 
how .align works on standard djgpp as I would appreciate the info as well.

Andrew

- Raw text -