This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: forwarded message from H. J. Lu
- From: "H. J. Lu" <hjl at lucon dot org>
- To: John David Anglin <dave at hiauly1 dot hia dot nrc dot ca>
- Cc: amodra at bigpond dot net dot au, doko at cs dot tu-berlin dot de, troup at debian dot org,bug-binutils at gnu dot org, binutils at sources dot redhat dot com
- Date: Tue, 20 May 2003 11:58:21 -0700
- Subject: Re: forwarded message from H. J. Lu
- References: <20030520103357.A7939@lucon.org> <200305201832.h4KIWxiJ023673@hiauly1.hia.nrc.ca>
On Tue, May 20, 2003 at 02:32:58PM -0400, John David Anglin wrote:
> > Follow
> >
> > http://sources.redhat.com/ml/binutils/2003-04/msg00403.html
> >
> > and all the issues it caused.
>
> As I'm sure you know, labels on the PA have to start at the beginning
> of a new line. They may or may not end with a colon. There has to be
> whitespace between the label, or the start of the line if there is no
> label, and the opcode. The opcode may contain completers. These are
> separated from the instruction mneumonic by commas. The opcode and the
> first operand are separated by whitespace. Operands are separated by
> a comma. A comment may follow the operand starting with a semicolon.
> A new statement can be started with '!' but it can't contain a label.
>
> It would appear safe to remove whitespace after commas in the opcode
> and operand fields. The code in tc-hppa.c assumes whitespace has
> been removed from the operands and probably from the opcode.
>
> I still don't fully understand what caused the breakage. It could be
> related to the ia64 label/opcode identification problem.
We no longer blindly remove whitespaces. Each backend which assumes
whitespace removed may be affected. When it happens, we will add a
testcase and fix it.
H.J.