From:	=?ISO-8859-1?Q?K=F6teles_K=E1roly?= <KOTELES AT phys DOT szote DOT u-szeged DOT hu>
Organization:	=?ISO-8859-1?Q?Albert_Szent-Gy=F6rgyi_Medical_Univ.?=
To:	djgpp AT delorie DOT com
Date:	Sun, 10 Oct 1999 20:44:27 MET
MIME-Version:	1.0
Subject:	as problem with sections and 8 bit relative addresses
X-Confirm-Reading-To:	=?ISO-8859-1?Q?K=F6teles_K=E1roly?= <KOTELES AT phys DOT szote DOT u-szeged DOT hu>
X-pmrqc:	1
X-mailer:	Pegasus Mail v3.31
Message-ID:	<DF65242527@phys.szote.u-szeged.hu>
Reply-To:	djgpp AT delorie DOT com

Dear DJGPP users,

 I'm using DJGPP 2.02 with gcc 2.8.1/binutils 2.8.1 and recently met with a 
problem. It appears only 
1. in functions put in a section different from ".text"
2. in inline asm
3. with instructions having 8 bit relative addresses (loop, jcxz)
The error message comes from the assembler and reads:
"Error: Value of .... too large for field of 1 bytes at 0x..."

I don't know if this is the appropriate mailing list; if not, sorry for 
disturbing you.

A few additional info: 1. everything works fine without the section 
attributes, 2. i couldn't check if this condition is necessary, gcc 
apparently doesn't emit loop/jcxz instructions (?), 3. insns having 
rel16/rel32 alternatives (all conditional jumps except jcxz) assemble.
The jumps are mostly to local labels only a few bytes away (but the 
problem exists with global labels as well). It seems to me as if the labels 
were defined being in the ".text", but the actual jump instructions in the 
other section.

The source and linker script fragments have been left out intentionally, i 
am only interested in if this is a known "issue"; i don't want to burden 
anyone with my buggy code :).

Thank you very much in advance!

=========================================
Koteles, Karoly
Dept. of Physiology
Albert Szent-Gyorgyi Medical University
Szeged, HUNGARY

koteles AT phys DOT szote DOT u-szeged DOT hu

- Raw text -